Sound level estimation

Abstract

Disclosed is a method of determining in real-time an estimate of a Sound Pressure Level or Sound Exposure corresponding to sound produced in an acoustic environment by multiple loudspeakers of a sound system in response to an input signal to the sound system. The method comprises determining in real-time the estimate of the Sound Pressure Level or Sound Exposure on the basis of the input signal. The sound system is a level-calibrated sound system. Also disclosed is a real-time sound level estimator. Also disclosed are a method and a real-time sound level estimator where the estimation is based on an approximation of system properties based on microphone measurements, or established for multiple listening positions, respectively. Also disclosed are methods of calibrating and monitoring a sound system and calibration and monitoring devices therefore.

Claims

1. A method of providing an estimate of a Sound Pressure Level (SPL) or Sound Exposure (SE) of sound produced in an acoustic environment (AE) by multiple loudspeakers of a sound system in response to an electrical or optical representation of an input audio signal, the acoustic environment having at least one listening position, the method comprising: an initial step of obtaining an approximation of sound system properties for each of the at least one listening position in the acoustic environment, the approximation of sound system properties being determined using a microphone measurement; receiving the electrical or optical input audio signal in a sound level detector/integrator (SSLD) in the sound system without using a microphone; determining in real time with the SSLD, in response to receiving the input audio signal and the approximation of sound system properties, individual estimates of the Sound Pressure Level (SPL-EST) or Sound Exposure (SE-EST) simultaneously at one or more of the at least one listening positions in the acoustic environment; and providing the determined SPL-EST or SE-EST as an output, wherein: the determining of SPL EST includes: (a) frequency weighting; (b) RMS integration based on: $x_{\mathrm{RMS}}\left(c,t\right)=\sqrt{\frac{1}{T}\underset{i=1}{\overset{T}{.Math.}}{x_A\left(c,t-i\right)}^2};$ (c) channel summation in response to a multi-channel audio input signal, otherwise omitting step (c) in response to receipt of a single channel audio input single; (d) establish a simulated averaged sound pressure in the AE; and (e) L.sub.eq calculation to generate the estimated, equivalent continuous sound level SPL_EST.

2. The method of claim 1, wherein the method comprises calculating an inter-channel correlation of one or more pairs of channels of the input signal, and further modifying the estimate of a Sound Pressure Level or Sound Exposure by using the calculated inter-channel correlation in the determining in real time the estimate of the Sound Pressure Level or Sound Exposure.

3. The method of claim 2, wherein the method comprises providing a pre-determined diffuseness of the acoustic environment, and further modifying the estimate of Sound Pressure Level or Sound Exposure by using the calculated inter-channel correlation and the diffuseness of the acoustic environment in the determining in real-time estimate of the Sound Pressure Level or Sound Exposure.

4. The method of claim 1, wherein the method comprises attenuating the sound produced by the sound system based on the estimate of Sound Pressure Level or Sound Exposure in order to limit the estimated Sound Pressure Level or Sound Exposure from exceeding a threshold.

5. The method of claim 1, wherein the approximation of sound system properties comprises a representation of a relation between an acoustic output level and an electrical or digital input level of the sound system in the acoustic environment.

6. The method of claim 1, wherein the approximation of sound system properties comprises approximations of frequency response of one or more of the multiple loudspeakers in the acoustic environment.

7. The method of claim 1, wherein the approximation of the sound system properties comprises approximations of reverberation characteristic of the acoustic environment.

8. The method of claim 1, wherein the approximation of sound system properties comprises a transfer function between each pair of one of the multiple loudspeakers and one of the at least one listening positions in the acoustic environment.

9. The method of claim 1, wherein the providing of the determined estimates as an output comprises visually displaying the determined estimates.

10. The method of claim 1, wherein the providing of the determined estimates as an output comprises logging the determined estimates.

11. The method of claim 1, wherein the providing of the determined estimates as an output comprises storing the determined estimates in a memory.

12. The method of claim 1, wherein the providing of the determined estimates as an output comprises transmitting the determined estimates over a network.

13. A method of providing an estimate of a Sound Pressure Level (SPL) or Sound Exposure (SE) of sound produced in an acoustic environment (AE) by multiple loudspeakers of a sound system in response to an electrical or optical representation of an input audio signal, the acoustic environment having at least one listening position, the method comprising: an initial step of obtaining an approximation of sound system properties for each of the at least one listening position in the acoustic environment, the approximation of sound system properties being determined using a microphone measurement; receiving the electrical or optical input audio signal in a sound level detector/integrator (SSLD) in the sound system without using a microphone; determining in real time with the SSLD, in response to receiving the input audio signal and the approximation of sound system properties, individual estimates of the Sound Pressure Level (SPL_EST) or Sound Exposure (SE_EST) simultaneously at one or more of the at least one listening positions in the acoustic environment; and providing the determined SPL_EST or SE_EST as an output, wherein: the determining of SE_EST includes: (a) frequency weighting; (b) RMS integration based on: $x_{\mathrm{INT}}\left(c,t\right)=\sqrt{\underset{i=1}{\overset{t}{.Math.}}{x_A\left(c,t-i\right)}^2};$ (c) channel summation in response to a multi-channel audio input signal, otherwise omitting step (c) in response to receipt of a single channel audio input single; (d) establish a simulated averaged sound pressure in the AE; and (e) L.sub.eq calculation to generate the estimated sound exposure level over a period of time SE_EST.

14. The method of claim 13, wherein the method comprises calculating an inter-channel correlation of one or more pairs of channels of the input signal, and further modifying the estimate of a Sound Pressure Level or Sound Exposure by using the calculated inter-channel correlation in the determining in real time the estimate of the Sound Pressure Level or Sound Exposure.

15. The method of claim 14, wherein the method comprises providing a pre-determined diffuseness of the acoustic environment, and further modifying the estimate of Sound Pressure Level or Sound Exposure by using the calculated inter-channel correlation and the diffuseness of the acoustic environment in the determining in real-time estimate of the Sound Pressure Level or Sound Exposure.

16. The method of claim 13, wherein the method comprises attenuating the sound produced by the sound system based on the estimate of Sound Pressure Level or Sound Exposure in order to limit the estimated Sound Pressure Level or Sound Exposure from exceeding a threshold.

17. The method of claim 13, wherein the approximation of sound system properties comprises a representation of a relation between an acoustic output level and an electrical or digital input level of the sound system in the acoustic environment.

18. The method of claim 13, wherein the approximation of sound system properties comprises approximations of frequency response of one or more of the multiple loudspeakers in the acoustic environment.

19. The method of claim 13, wherein the approximation of the sound system properties comprises approximations of reverberation characteristic of the acoustic environment.

20. The method of claim 13, wherein the approximation of sound system properties comprises a transfer function between each pair of one of the multiple loudspeakers and one of the at least one listening positions in the acoustic environment.

21. The method of claim 13, wherein the providing of the determined estimates as an output comprises visually displaying the determined estimates.

22. The method of claim 13, wherein the providing of the determined estimates as an output comprises logging the determined estimates.

23. The method of claim 13, wherein the providing of the determined estimates as an output comprises storing the determined estimates in a memory.

24. The method of claim 13, wherein the providing of the determined estimates as an output comprises transmitting the determined estimates over a network.

25. A real-time sound level estimator for real-time estimation of Sound Pressure Level (SPL) or Sound Exposure (SE) of sound produced in an acoustic environment (AE) by multiple loudspeakers of a sound system in response to an electrical or optical representation of an input audio signal, the acoustic environment having at least one listening position, the sound level estimator comprising: an input; a processor coupled to the input; a memory accessible by the processor, the memory structured to store instructions for the processor; and an output coupled to the processor; wherein the processor is arranged to perform the following steps: an initial step of obtaining an approximation of sound system properties for each of the at least one listening position in the acoustic environment, the approximation of sound system properties being determined using a microphone measurement; receiving the electrical or optical input audio signal in a sound level detector/integrator (SSLD) in the sound system without using a microphone; determining in real time with the SSLD, in response to receiving the input audio signal and the approximation of sound system properties, individual estimates of the Sound Pressure Level (SPL_EST) or Sound Exposure (SE_EST) simultaneously at one or more of the at least one listening positions in the acoustic environment; and providing the determined SPL_EST or SE_EST as an output, wherein: the determining of SPL_EST includes: (a) frequency weighting; (b) RMS integration based on: $x_{\mathrm{RMS}}\left(c,t\right)=\sqrt{\frac{1}{T}\underset{i=1}{\overset{T}{.Math.}}{x_A\left(c,t-i\right)}^2};$ (c) channel summation in response to a multi-channel audio input signal, otherwise proceeding to step (d) in response to receipt of a single channel audio input single; (d) establish a simulated averaged sound pressure in the AE; and (e) L.sub.eq calculation to generate the estimated, equivalent continuous sound level SPL_EST.

26. A real-time sound level estimator for real-time estimation of Sound Pressure Level (SPL) or Sound Exposure (SE) of sound produced in an acoustic environment (AE) by multiple loudspeakers of a sound system in response to an electrical or optical representation of an input audio signal, the acoustic environment having at least one listening position, the sound level estimator comprising: an input; a processor coupled to the input; a memory accessible by the processor, the memory structured to store instructions for the processor; and an output coupled to the processor; wherein the processor is arranged to perform the following steps: an initial step of obtaining an approximation of sound system properties for each of the at least one listening position in the acoustic environment, the approximation of sound system properties being determined using a microphone measurement; receiving the electrical or optical input audio signal in a sound level detector/integrator (SSLD) in the sound system without using a microphone; determining in real time with the SSLD, in response to receiving the input audio signal and the approximation of sound system properties, individual estimates of the Sound Pressure Level (SPL_EST) or Sound Exposure (SE_EST) simultaneously at one or more of the at least one listening positions in the acoustic environment; and providing the determined SPL_EST or SE_EST as an output, wherein: the determining of SE_EST includes: (a) frequency weighting; (b) RMS integration based on: $x_{\mathrm{INT}}\left(c,t\right)=\sqrt{\underset{i=1}{\overset{t}{.Math.}}{x_A\left(c,t-i\right)}^2};$ (c) channel summation in response to a multi-channel audio input signal, otherwise proceeding to step (d) in response to receipt of a single channel audio input single; (d) establish a simulated averaged sound pressure in the AE; and (e) L.sub.eq calculation to generate the estimated sound exposure level over a period of time SE_EST.

Description

DRAWINGS

(1) Various embodiments of the invention will in the following be described with reference to the drawings where

(2) FIGS. 1-3 illustrate prior art principles for measuring or controlling sound level,

(3) FIG. 4 illustrates an embodiment of the invention based on a level-calibrated sound system,

(4) FIG. 5 illustrates a frequency response of an A-weighting filter,

(5) FIG. 6 illustrates an embodiment of the invention comprising an inter-channel correlation estimator,

(6) FIG. 7 shows a plot from an experiment with inter-channel correlation estimation according to an embodiment,

(7) FIG. 8 illustrates an embodiment of the invention comprising transfer function characteristics and a transfer filter,

(8) FIG. 9 illustrates an embodiment of the invention comprising transfer function characteristics and a transfer filter to produce estimation for several listening positions,

(9) FIG. 10 illustrates an embodiment of the invention comprising frequency response characteristics and a frequency weighting,

(10) FIG. 11 shows a plot of a frequency response from an experiment,

(11) FIG. 12 shows a spectrogram of a test signal used for an experiment,

(12) FIG. 13 shows the result of an experiment of estimating an L.sub.Aeq,1 representation of SPL computed by 3 different embodiments,

(13) FIG. 14 illustrates an embodiment of the invention where approximated system properties are based on microphone measurements,

(14) FIG. 15 illustrates a first step of an embodiment of a method of the invention for obtaining approximated system properties,

(15) FIG. 16 illustrates a second step of an embodiment of a method of the invention for using obtained approximated system properties in estimating SPL and SE,

(16) FIG. 17 illustrates an embodiment of the invention employing an acoustic space emulator, and

(17) FIG. 18 illustrates an embodiment of the invention employing an acoustic space SPL emulator.

DETAILED DESCRIPTION OF THE EXAMPLE EMBODIMENTS

(18) It is noted that an audio amplifier, as shown e.g. in FIG. 4, FIG. 9 and FIGS. 14-16, though typically part of a sound system and included in preferred embodiments of the invention, may not be implemented in some embodiments of the invention. Hence, an estimate of acoustic properties, preferably a Sound Pressure Level or Sound Exposure, determined by various embodiments of the present invention, may take place before, after, or integrated with the amplification of the input signal, prior to rendering by the loudspeakers.

(19) FIG. 4 is a diagram showing a sound system SS which receives an input signal IS and produces sound in an acoustic environment AE based on the input signal. The sound system comprises an audio amplifier AA and multiple loudspeakers LS. The sound system may optionally comprise further audio processing modules, e.g. decoder, equalizer, compressor, filter, etc. The input signal is a mono-, stereo- or multichannel analog or digital audio signal and the sound system may receive the input signal from an external source or itself integrate with a radio, media player, mixer table, or other audio source establishing the input signal.

(20) The audio amplifier may be any kind of audio amplifier, and the number of channels should preferably correspond to the relevant loudspeaker setup. The multiple loudspeakers are located at different positions relative to the acoustic environment, and may produce audio from different audio channels and/or from common audio channels. For example, the audio amplifier may be a stereo amplifier and the multiple loudspeakers may comprise two speakers receiving different audio channels in a stereo setup or e.g. the audio amplifier is a multi-channel receiver and multiple loudspeakers comprise 6 speakers receiving different audio channels in a surround sound setup, including a subwoofer channel. In another example, the multiple loudspeakers may comprise 25 speakers all rendering the same audio channel, for example for background music in a supermarket. In another example, the multiple loudspeakers may comprise 8 speakers whereof three speakers are positioned along a left side of a room and all receiving a left channel, three speakers likewise positioned along a right side of the room and all receiving a right channel, and the remaining two speakers receiving a subwoofer channel.

(21) Further, a Mic-less sound level estimator MLE is provided, which is arranged, as described in further detail below, to provide, on the basis of the input signal, an estimate EST of the SPL or SE produced in at least one position in the acoustic environment. In the present embodiment, the Mic-less sound level estimator consists of a Simulating sound level detector and integrator SSLD.

(22) In the embodiment of FIG. 4 the sound system SS is level-calibrated with its loudspeakers LS in the acoustic environment AE, such that the relation between levels of the input signal IS and the resulting Sound-Pressure Levels in the acoustic environment AE is known. The level calibration may for example be achieved by measuring the sound pressure level produced by a calibration signal consisting of pink noise at an RMS-level of 20 dBFS, for example in a sound system with the input signal IS being a digital stereo signal (i.e. 2 channels). For simplicity of this description, frequency-weighting in the level-calibration is ignored, without limiting the scope of the present invention. This calibration signal, played through either the left or the right input signal channel, through a stereo sound system, will then produce, for example, 80 dB SPL in the acoustic environment AE. This is what the level-calibration asserts, for at least one listening-position in the acoustic environment AE. In practice, there may be many different methods and official recommendations for achieving the level-calibration, as known by the person skilled in the art.

(23) Given the level-calibration, the Simulating sound level detector and integrator SSLD can calculate estimates EST such as Sound Pressure Level estimates SPL-EST and Sound Exposure estimates SE-EST, in real-time and on the basis of the input signal, as follows, with the procedure applying equally to other calibration-signals, and -levels:

(24) Procedure to Compute an SPL-EST of the L.sub.Aeq,15 Measure, Formally Known as the Equivalent-Continuous A-Weighted Sound Level, 15 Minute Time Period:

(25) This description and the following is based on discrete-time processing, as would be employed in digital implementation of the embodiments. In this case, the input signal IS would be in a digital format, or it would be converted to digital by an A/D converter. The processing steps could be implemented on a DSP or a CPU with support from relevant kinds of memory. The resulting SPL-EST or SE-EST could for example be stored in memory, logged for future reference, transmitted over a network, and/or shown on a display.

(26) Step 1. A-Weighting

(27) Each individual channel of the input signal IS is filtered with an A-weighting filter. The filter is designed to match the requirements for the A frequency weighting, as defined in International standard IEC 61672 and ANSI S 1.4. A frequency response AW of an A-weighting filter is illustrated in FIG. 5.

(28) The operation is shown in Eq. 1 as a convolution, with A(i) being the i'th coefficient of a direct form FIR-filter implementing the A-weighting, with A.sub.n being the filter order. The x(c,t) refers to channel c of the input signal IS at time t. In an advantageous embodiment, an IIR filter could be used to implement the A-weighting filter, as a computational optimization. Here and in the rest of the description, a separate channel typically also means a separate loudspeaker from the multiple loudspeakers. In other words, the number of audio channels and the number of loudspeakers are preferably equal.

(29) $\begin{array}{cc}{x}_A\left(c,t\right)=\underset{i=1}{\overset{A_n}{.Math.}}x\left(c,t-i\right).Math.A\left(i\right)& \left(\mathrm{Eq}.1\right)\end{array}$

(30) Step 2. RMS-Integration

(31) Defining T as the number of samples in the input signal IS corresponding to the time period of the L.sub.eq-measure in question, a sampling rate of 48.0 kHz, results in T=15 min=43.2 MS (mega-samples, i.e. million samples). Integration of the values of each channel may be performed according to the formula Eq. 2 below. In an advantageous embodiment, the sum in Eq. 2 is split into several buffered sub-sums over smaller periods, corresponding to the necessary time-resolution of the estimate, as a computational optimization.

(32) $\begin{array}{cc}{x}_{\mathrm{RMS}}\left(c,t\right)=\sqrt{\frac{1}{T}\underset{i=1}{\overset{T}{.Math.}}{x_A\left(c,t-i\right)}^2}& \left(\mathrm{Eq}.2\right)\end{array}$

(33) Step 3. Channel-Summation

(34) As the channels are mixed in the acoustic environment, their individual values have to be combined, which may preferably be performed according to Eq. 3. Here C.sub.n is the number of channels of input signal IS (e.g. C.sub.n=2, for stereo). Essentially, X.sub.SUM is the total, averaged energy of the frequency-weighted input signal, over the most recent time-period of the given duration.

(35) $\begin{array}{cc}{x}_{\mathrm{SUM}}\left(t\right)=\sqrt{\underset{c=1}{\overset{C_n}{.Math.}}{x_{\mathrm{RMS}}\left(c,t\right)}^2}& \left(\mathrm{Eq}.3\right)\end{array}$

(36) Step 4. Simulated Sound Pressure

(37) To establish a simulated, averaged sound pressure in the acoustic environment in units of Pascal (Pa), the calculated averaged signal energy from Eq. 3 is combined with the result of the level calibration described above, by Eq. 4. L.sub.ref corresponds to the level of the calibration signal, for example 20 dBFS as in the above example, and L.sub.cal corresponds to the Sound Pressure Level produced by the sound system SS in response to the calibration signal, for example 80 dB SPL as was measured in the above example. In Eq. 4 is also applied that a sound pressure of 1 Pa corresponds to 94.0 dB SPL.
p.sub.SUM(t)={square root over (x.sub.SUM(t).sup.2.Math.10.sup.(L.sup.ref.sup.+(L.sup.cal.sup.94))/10)}Pa(eq. 4)

(38) Step 5. L.sub.eq-Calculation

(39) An estimated, equivalent-continuous sound level (SPL_EST) can be computed by Eq. 5 based on the above results. In Eq. 5 the standard reference sound pressure, p.sub.0=20 Pa, is used.

(40) $\begin{array}{cc}{L}_{\mathrm{Aeq},15}\left(t\right)=10.Math.{\log }_{10}\left(\frac{p_{\mathrm{SUM}}^2\left(t\right)}{p_0^2}\right)\mathrm{dB}& \left(\mathrm{Eq}.5\right)\end{array}$

(41) Procedure to Compute an SE-EST of the L.sub.EX,8h, Formally Known as the Noise Exposure Level Normalized to a Nominal 8 h Working Day:

(42) Eq. 2 is replaced with Eq. 6, calculating an integration instead of an average. In practice, this integrator may be reset (i.e. t:=0), for example, at the beginning of each working day.

(43) $\begin{array}{cc}{x}_{\mathrm{INT}}\left(c,t\right)=\sqrt{\underset{i=1}{\overset{t}{.Math.}}{x_A\left(c,t-i\right)}^2}& \left(\mathrm{Eq}.6\right)\end{array}$

(44) Eq. 3-4 are employed unchanged, i.e. employing x.sub.INT instead of x.sub.RMS, leading to the final calculation of L.sub.EX,8h. Here T.sub.n=8 hours, as the standard working day duration, based on which the sound exposure is normalized. Other measures of sound exposure (noise exposure), such as E, SEL, L.sub.avg, and Dose %, can be calculated analogously.

(45) $\begin{array}{cc}{L}_{\mathrm{ex},8h}\left(t\right)=10.Math.{\log }_{10}\left(\frac{p_{\mathrm{SUM}}^2\left(t\right)}{p_0^2}\right)+10.Math.{\log }_{10}\left(\frac{T}{T_n}\right)\mathrm{dB}& \left(\mathrm{Eq}.7\right)\end{array}$

(46) Procedure to Compute an SPL-EST of the L.sub.Cpeak Measure, Formally Known as the Peak Sound Pressure Level, C-weighted:

(47) Some regulations require the absolute peak SPL to be measuredin addition to an L.sub.eqto protect against hearing damages caused by sudden large pressure peaks and high-level transient noises.

(48) Step 1. C-Weighting

(49) The C-weighting is commonly employed in connection with peak SPL measurement. It has a different frequency response than the A-weighting, but may be implemented similarly to Eq. 1. The x(c,t) refers to channel c of the input signal IS at time t.

(50) $\begin{array}{cc}{x}_C\left(c,t\right)=\underset{i=1}{\overset{C_n}{.Math.}}x\left(c,t-i\right).Math.C\left(i\right)& \left(\mathrm{Eq}.8\right)\end{array}$

(51) Step 2. Peak Level Detector.

(52) The absolute peak sound level is estimated, in the input signal-domain.

(53) $\begin{array}{cc}{x}_{\mathrm{PK}}\left(c,t\right)=\underset{1iT}{\max }.Math.x_C\left(c,t-i\right).Math.& \left(\mathrm{Eq}.9\right)\end{array}$

(54) Step 3. Channel-Summation

(55) The peak level is estimated across channels of the input signal IS.

(56) $\begin{array}{cc}{x}_{\mathrm{SUM}}\left(t\right)=\underset{1c{C}_n}{\max }{x}_{\mathrm{PK}}\left(c,t\right)& \left(\mathrm{Eq}.10\right)\end{array}$

(57) Step 4. Simulated Sound Pressure

(58) Analogous to Eq. 4.
p.sub.SUM(t)={square root over (x.sub.SUM(t).sup.2.Math.10.sup.(L.sup.ref.sup.+(L.sup.cal.sup.94))/10)}Pa(Eq. 11)

(59) Step 5. L.sub.peak Calculation

(60) The estimated, peak sound level can now be computed. Equivalently, the parameter p.sub.peak would be L.sub.peak in units of Pascal.

(61) 0$\begin{array}{cc}{L}_{C,\mathrm{peak}}\left(t\right)=10.Math.{\log }_{10}\left(\frac{p_{\mathrm{SUM}}^2\left(t\right)}{p_0^2}\right)\mathrm{dB}& \left(\mathrm{Eq}.12\right)\end{array}$

(62) Inter-Channel Correlation

(63) The accuracy of the channel-sum estimate, calculated in Eq. 3, may be improved by taking into account an inter-channel correlation.

(64) FIG. 6 shows a diagram of an embodiment with an input signal IS feeding a sound system SS consisting of an amplifier connected to multiple loudspeakers which produce sound in an acoustic environment AE; the sound system SS has been level-calibrated in the acoustic environment AE. Also fed by the input signal is a Mic-less sound level estimator MLE, consisting of a Simulating sound level detector SSLD, capable of producing estimates of SPL and/or Sound Exposure SPL EST. The embodiment of FIG. 6 resembles that of FIG. 4 with the addition that the Mic-less sound level estimator MLE in the present embodiment further comprises an Inter-channel correlation estimator ICCE, which also receives the input signal IS.

(65) Thereby this embodiment is enabled to take into consideration that two or more audio channels may be correlated, and thereby cause a real sound pressure which is higher than the estimated SPL. The error of an estimation not taking inter-channel correlation into account may be as high as 3 dB for 2 channels, and even higher for more correlated channels, for example 7 dB for 5 channels. Hence, it is preferred to provide embodiments of the invention with inter-channel correlation estimation. The inter-channel correlation estimator is preferably implemented in the same DSP or CPU implementing the Simulating sound level detector and integrator SSLD, and also receives the input signal IS.

(66) Procedure to Compute an SPL_EST Employing the Inter-Channel Correlation Estimator ICCE:

(67) One embodiment of Mic-less sound level estimator implementing inter-channel correlation estimation is described here. The example assumes a stereo input signal, e.g. one pair of channels c.sub.1 and c.sub.2, and a selected time frame for the correlation calculation of T.sub.short=50 ms (in samples), for example. T.sub.short may preferably be selected so that it roughly corresponds to the period of the lowest significant frequency produced by the sound system SS.

(68) The inter-channel correlation, at time t, for channel-pair c.sub.1 and c.sub.2, is denoted r(t,c.sub.1,c.sub.2), and is estimated by the inter-channel correlation estimator ICCE. Eq. 13 calculates the sample correlation coefficient for the two discrete-time signals x(c.sub.1,t) and x(c.sub.2,t), as a function of time. In Eq. 13, the time indices on x(c,t) and on the 's have been omitted, for clarity; these terms always relate to the range of most recent T.sub.short samples, for both channels.

(69) $\begin{array}{cc}r\left(t,c_1,c_2\right)=\frac{T_{\mathrm{short}}.Math.x\left(c_1\right)x\left(c_2\right)-.Math.x\left(c_1\right).Math.x\left(c_2\right)}{\begin{array}{c}\sqrt{T_{\mathrm{short}}.Math.{\left(x\left(c_1\right)\right)}^2-{\left(.Math.x\left(c_1\right)\right)}^2}\\ \sqrt{T_{\mathrm{short}}{\left(x\left(c_2\right)\right)}^2-{\left(.Math.x\left(c_2\right)\right)}^2}\end{array}}& \left(\mathrm{Eq}.13\right)\end{array}$

(70) The sound level detector and integrator SSLD may be the same as described above with reference to FIG. 4, except for the following changes to Eq. 2 and 3, based on the the r( )-function of Eq. 13; thereby this embodiment of the sound level detector and integrator SSLD adapts its estimate to properties in addition to the time-varying level of the input signal channels.

(71) $\begin{array}{cc}{x}_{\mathrm{RMS}}\left(c,t\right)=\sqrt{\frac{1}{T}\underset{i=1}{\overset{T_{\mathrm{short}}}{.Math.}}{x_A\left(c,t-i\right)}^2}\mathrm{and}& \left(\mathrm{Eq}.14\right)\\ x_{\mathrm{SUM}}\left(t\right)=\sqrt[q\left(t,c_1,c_2\right)]{\underset{i=1}{\overset{T/T_{\mathrm{short}}}{.Math.}}{x_{\mathrm{RMS}}\left(c_1,t\right)}^{q\left(t,c_1,c_2\right)}+{x_{\mathrm{RMS}}\left(c_2,t\right)}^{q\left(t,c_1,c_2\right)}}& \left(\mathrm{Eq}.15\right)\end{array}$

(72) where q is in the interval [1,2], and may be defined as:
q(t,c1,c2)=2max(0,r(t,c1,c2))(Eq. 16)

(73) The max( )-operator returns the maximum value of 0 and the r( )-function, as the effect of destructive interference of channels may not be desirable to include due to its dependency on frequency and specific position in the acoustic environment AE. Other embodiments may consider destructive interference to some degree.

(74) Other q( )-functions may be implemented in various embodiments with various effects and advantages, some of which are applied below, and may be applied in this embodiment instead of Eq. 16, mutatis mutandis.

(75) FIG. 7 shows a plot of an example of the consequences of inter-channel correlation of a 2-channel input signal IS consisting of 3 sound segments, 10 s each.

(76) The first segment SEG1 from 0 to 10 seconds contains pink noise IS1, based on noise generated independently for the 2 channels (i.e. inter-channel correlation of 0.0).

(77) The second segment SEG2 from 10 to 20 seconds contains pink noise IS2, based on noise generated specifically to have an inter-channel correlation of 0.66.

(78) The third segment SEG3 from 20 to 30 seconds contains pink noise IS3, same generated signal for the 2 channels (i.e. inter-channel correlation of 1.0).

(79) Note that the 3 segments in the experiment have identical levels and frequency spectrums, both within and across the individual channels; that is, the interchannel-correlation is the only difference. The plot shows the actual waveform, the corresponding values r1, r2, r3, q1, q2, q3, of the r( ) and q( ) functions as calculated from Eq. 13 and 16, and the resulting x.sub.SUM( ) SUM1, SUM2, SUM3, of Eq. 15 over time.

(80) From the experiment first segment SEG1 it can be seen, that an embodiment failing to take the inter-channel correlation into account, would normally produce an x.sub.SUMSUM1 around 11 dB (in the example), i.e. corresponding to a correlation R1=r(t)=0. On the other hand, for the segment SEG3 from 20 to 30 seconds with complete correlation, i.e. R3=r(t)=1, the resulting x.sub.SUM is around 8 dB. Hence, an error of up to 3 dB (in the experiment) would be made by not considering inter-channel correlation in the SPL estimation.

(81) Another example of an embodiment of a sound system has 4 channels, where the 4-channel input signal IS of the example is known to comprise inter-channel correlation within two channel-pairs, but not between the other channels; that is c.sub.1 and c.sub.2 may be correlated, and c.sub.3 and c.sub.4 may be correlated. In this case, Eq. 15 is applied three times, with three different q( ) functions:
q(t,c1,c2)=2max(0,r(t,c1,c2))(Eq. 17)
q(t,c3,c4)=2max(0,r(t,c3,c4))(Eq. 18)
and
q(t,c12,c34)=2(Eq. 19)

(82) where Eq. 17 and 18 correspond to Eq. 16 and relates to two channels, each, and Eq. 19 is for summing the two channel-pairs together to total, under the assumption that the pairs are not correlated. Eq. 19 may be adjusted to include a similar r ( )-function if correlation between the pairs could also be expected. In an embodiment, the q( )-functions may be dynamically implemented so that they may relate to the channels pairs showing the highest correlation at a specific point in time, thereby for example calculating correlation for all channel pairs, but only applying Eq. 15 for the channels pairs showing a correlation above a predetermined threshold, e.g. 0.5, and summing the remaining channels according to Eq. 3. In another embodiment, particularly relevant for sound system SS with a high number of channels, e.g. 22 channels, it may be advantageous to perform down-mixes of a number of adjacent channels, e.g. each 3 channels, and then calculate correlation between pairs of downmixes.

(83) In another embodiment with a larger number of channels, e.g. PA systems, loudspeaker installations in a building, etc. it may be advantageous to calculate the level of one or more sub-mixes, that is, first summing the signals in certain subsets of the input signal IS channels. These channel subsets may be selected such that the contributions from adjacent loudspeakers are summed, and in the subsequent computations considered as one signal.

(84) Furthermore, the level of such a sub-mix and the channel-summed individual levels (e.g. Eq. 3) may be employed as a high and low bound of the estimate, respectively. Hence, the estimated channel-summed level may continually be chosen somewhere in between these bounds; this choice could be refined further by incorporating the diffuseness of the room and/or the proximity of the loudspeaker involved. This method may lead to a more accurate estimate of SPL and/or SE, e.g. for a larger number of loudspeakers or as a supplement to the inter-channel correlation estimator ICCE method.

(85) In an advanced embodiment, the SPL_EST and SE_EST estimates may be improved further by taking into account that the significance of inter-channel correlation in real world setups depends on the diffuseness of the acoustic environment AE. In an embodiment the diffuseness is predetermined, e.g. by assumption by a user or from room specifications, and input as a number, D, which denotes the diffuseness as a continuous parameter in the interval [0,1], where, 0=very damped acoustic environment AE (e.g. an anechoic chamber), aka. free field, and 1=very reverberant, aka. diffuse field. The q( )-function of Eq. 16 is substituted with Eq. 20 to take into account diffuseness in the handling of inter-channel correlation:
q(t,c1,c2)=max(2,2max(0,r(t,c1,c2))+D)(Eq. 20)

(86) In alternative embodiments the q( )-function may be implemented differently, e.g. by scaling instead of limiting in Eq. 20 to stay below a value of 2, by designing r( ) and D to be multiplicative instead of additive, etc.

(87) An advanced embodiment, estimating the L.sub.Cpeak, may also benefit from a pre-determined diffuseness D of the acoustic environment AE. In this case, Eq. 9 and Eq. 10 may be replaced with Eq. 21 to 23:
q=D+1(Eq. 21)

(88) where the instantaneous peak sound level, in the input signal domain, is estimated as:

(89) $\begin{array}{cc}{x}_{\mathrm{SUM}}\left(t\right)=\sqrt[q]{\underset{c=1}{\overset{C_n}{.Math.}}{.Math.x\left(c,t\right).Math.}^q}& \left(\mathrm{Eq}.22\right)\end{array}$

(90) and the the peak level, over time, would then be:

(91) $\begin{array}{cc}{x}_{\mathrm{PK}}\left(t\right)=\underset{1iT}{\max }{x}_{\mathrm{SUM}}\left(t-i\right)& \left(\mathrm{Eq}.23\right)\end{array}$

(92) As seen, the q( )-function handling the combining of contributions from different channels, may in some embodiments be independent of a channel correlation function r( ). In Eq. 21, a higher diffuseness, i.e. with D closer to 1.0, implies that the combined effect of individual channels on the total peak sound level is smaller, due to the (diffuse) reflections being relatively strong compared to the direct sound.

(93) Transfer Function Characteristics

(94) FIG. 8 shows a diagram of an embodiment with an input signal IS feeding a sound system SS consisting of an amplifier connected to multiple loudspeakers which produce sound in an acoustic environment AE; the Sound System has been level-calibrated in the acoustic environment. Also fed by the input signal is a Mic-less sound level estimator MLE, consisting of a Simulating sound level detector SSLD, capable of producing estimates EST of SPL and/or Sound Exposure SPL_EST. The embodiment of FIG. 8 resembles that of FIG. 4 with the addition that the Mic-less sound level estimator MLE in the present embodiment further comprises Transfer function characteristics TFC of loudspeakers in the acoustic environment AE and a transfer function filter TFF for applying the transfer function characteristics TFC to the input signal IS in connection with the estimation of SPL or SE.

(95) Thereby this embodiment is enabled to take into consideration the influence on the sound produced by the multiple loudspeakers, the room and furniture or other environment properties, preferably in consideration of a specific listening position. The transfer function characteristics may represent frequency response, phase response, gain, reflections, reverberation, etc. The estimated SPL for that specific setup and listening position may thereby be more accurate. The transfer function filter is preferably implemented in the same DSP or CPU implementing the simulating sound level detector and integrator SSLD, and also receives the input signal IS. The transfer function characteristics TFC may preferably be stored in memory included in or connected to the DSP or CPU.

(96) The transfer function characteristics TFC may be obtained by performing an impulse response measurement of the sound system SS in the acoustic environment AE, per channel, per listening position. Several ways of measuring such an impulse response are known in the art, e.g. by using pure tone frequency-sweeps (also known as chirps), typically in combination with a Fourier analysis with appropriate windowing and averaging properties. Equivalent to the impulse response would be the actual transfer function, which is the quotient of the cross power spectral density and the power spectral density between a test signal input as input signal IS and the corresponding acoustically measured signal. The transfer function characteristics TFC may be measured for example when the sound system SS is installed, or in connection with maintenance or re-calibration, or prior to a certain event or performance. In an embodiment, a test signal as described below with reference to FIG. 15 is used to determine the transfer function characteristics TFC.

(97) For each listening-position, each channel of the input signal may be processed by a filter implementing the transfer function for the corresponding loudspeaker (i.e. convolution). An improved estimate of SPL or Sound Exposure could be determined, by performing this processing, preferably as a first step, in determining said estimate. Advantageous of this embodiment is that it essentially considers the combined effects of the frequency response, the effect of inter-channel correlation including potentially both constructive and destructive interference, and the reverberation.

(98) Procedure to Compute an SPL_EST Employing the Transfer Function Characteristics TFC for a Single Listening Position:

(99) One embodiment of Mic-less sound level estimator MLE implementing transfer function characteristics TFC and corresponding filtering is described here, where x(c,t) refers to channel c of the input signal IS at time t. Further details of the variables and parameters may be found in the description of the embodiment of FIG. 4 above, mutatis mutandis.

(100) Step 1: Transfer Function Filter

(101) Let TF(c,i) denote the i'th coefficient of the transfer function for channel c from, and including, the loudspeaker to a specific listening position, realized as a FIR filter (direct form). Apply the FIR filter implementing the transfer function, per channel, to the input signal IS.

(102) $\begin{array}{cc}{x}_{\mathrm{TF}}\left(c,t\right)=\underset{i=1}{\overset{{\mathrm{TF}}_n}{.Math.}}x\left(c,t-i\right).Math.\mathrm{TF}\left(c,i\right)& \left(\mathrm{Eq}.24\right)\end{array}$

(103) As the transfer function may represent an impulse response of the acoustic environment AE, lasting several seconds to represent reverberation, etc., the order of the FIR filter, implementing the transfer function, may be 100,000 or more. In an advanced embodiment, the transfer function filter is implemented as FFT-based fast convolution, to reduce the computational complexity.

(104) Step 2: Channel-Summation

(105) The channels of the input signal IS are mixed. Note that this sample-level type of channel summation is appropriate in preferred embodiments where the TF(c) represents time- and frequency-domain characteristics for each channel; as opposed to other embodiments without the transfer function, where an energy summation of the channels would be preferable.

(106) $\begin{array}{cc}{x}_{\mathrm{MIX}}\left(t\right)=\underset{c=1}{\overset{C_n}{.Math.}}{x}_{\mathrm{TF}}\left(c,t\right)& \left(\mathrm{Eq}.25\right)\end{array}$

(107) Step 3. A-Weighting

(108) The A-weighting is the same for all channels, and thus the A-weighting filter may advantageously be applied after the channel-summation. Further details regarding A-weighting is given above with reference to FIG. 5.

(109) $\begin{array}{cc}{x}_A\left(t\right)=\underset{i=1}{\overset{A_n}{.Math.}}{x}_{\mathrm{MIX}}\left(t-i\right).Math.A\left(i\right)& \left(\mathrm{Eq}.26\right)\end{array}$

(110) Step 4. RMS-Integration.

(111) To make the filtered, mixed and A-weighted, but still sample-by-sample represented input signal more steady for the estimation, it is RMS-integrated, for example according to Eq. 27. This step could be replaced with an averaging, an integration, or a peak-detection, depending on the kind of SPL_EST or SE_EST in question.

(112) $\begin{array}{cc}{x}_{\mathrm{RMS}}\left(t\right)=\sqrt{\frac{1}{T}\underset{i=1}{\overset{T}{.Math.}}{x_A\left(t-i\right)}^2}& \left(\mathrm{Eq}.27\right)\end{array}$

(113) Step 5. Simulated Sound Pressure

(114) A simulated, averaged sound pressure in the acoustic environment in units of Pascal (Pa) is established by combining the obtained x.sub.RMS(t) with the level calibration of the system, as described above:
p.sub.SUM(t)={square root over (x.sub.RMS(t).sup.2.Math.10.sup.(L.sup.ref.sup.+(L.sup.cal.sup.94))/10)}Pa(Eq. 28)

(115) Step 6. L.sub.eq-Calculation

(116) Finally, an estimated, equivalent-continuous sound level (SPL_EST) can be computed by Eq. 5 above according to the corresponding description thereof, using p.sub.SUM(t) obtained by Eq. 28.

(117) If alternatively or additionally, an sound exposure SE-EST of the L.sub.EX,8h is desired, it may be calculated by substituting Eq. 27 with Eq. 29:

(118) $\begin{array}{cc}{x}_{\mathrm{RMS}}\left(t\right)=\sqrt{\underset{i=1}{\overset{t}{.Math.}}{x_A\left(t-i\right)}^2}& \left(\mathrm{Eq}.29\right)\end{array}$

(119) and using Eq. 7 for calculating the sound exposure from the p.sub.SUM(t) produced by Eq. 28.

(120) Likewise, other estimate types, e.g. the other types mentioned above, may be calculated for a single listening position based on transfer function characteristics and transfer function filter of the embodiment of FIG. 8, from the templates described above, mutatis mutandis.

(121) Multiple Listening Positions

(122) The embodiments described above have produced SPL or SE estimates EST relating to a single listening position. FIG. 9 presents an embodiment that enables the invention to apply to multiple listening-positions, each with an individual SPL or SE estimate EST. This extension may preferably be applied to any of the other embodiments, mutatis mutandis, when it may be relevant to monitor SPL or sound exposure in more positions simultaneously.

(123) FIG. 9 shows a diagram of an embodiment with an input signal IS feeding a sound system SS consisting of an amplifier AA connected to multiple loudspeakers LS which produce sound in an acoustic environment AE. Also fed by the input signal is a Mic-less sound level estimator MLE, comprising of a Simulating sound level detector SSLD, capable of producing estimates EST1, EST2, of SPL and/or Sound Exposure SPL EST for at least two positions, pos. I, LP1, and pos. II, LP2. The Mic-less sound level estimator MLE comprises characteristics LPC of positions I and II, for example transfer function characteristics from each loudspeaker to each listening position. The characteristics may be stored in a memory integrated with or connected to a DSP or CPU or other processing means implementing the sound level detector and integrator SSLD.

(124) Procedure to Compute an SPL_EST Employing the Transfer Function Characteristics TFC and the Multiple Listening-Positions Feature:

(125) Here follows an example embodiment of adjusting one of the above procedures to multiple listening positions according to the above.

(126) Step 1: Transfer Function Filter

(127) In the present embodiment, the characteristics LPC of the listening positions comprises transfer function characteristics TFC for each combination of loudspeaker and listening position. A FIR filter implementing the TF is applied for each listening-position 1p, for each channel c, i.e. each loudspeaker. For example, for 3 listening-positions and a sound system SS with 5 channels, a total of 15 transfer functions would need to be determined, and applied as filters. The following steps and equations resemble the above-described corresponding calculations with the addition of a listening position variable, and the above-described details, naming conventions and constraints or presumptions also apply to the below:

(128) 0$\begin{array}{cc}{x}_{\mathrm{TF}}\left(\mathrm{lp},c,t\right)=\underset{i=1}{\overset{{\mathrm{TF}}_n}{.Math.}}x\left(c,t-i\right).Math.\mathrm{TF}\left(\mathrm{lp},c,i\right)& \left(\mathrm{Eq}.30\right)\end{array}$

(129) Step 2: Channel-Summation

(130) $\begin{array}{cc}{x}_{\mathrm{MIX}}\left(\mathrm{lp},t\right)=\underset{c=1}{\overset{C_n}{.Math.}}{x}_{\mathrm{TF}}\left(\mathrm{lp},c,t\right)& \left(\mathrm{Eq}.31\right)\end{array}$

(131) Step 3. A-Weighting

(132) $\begin{array}{cc}{x}_A\left(\mathrm{lp},t\right)=\underset{i=1}{\overset{A_n}{.Math.}}{x}_{\mathrm{MIX}}\left(\mathrm{lp},t-i\right).Math.A\left(i\right)& \left(\mathrm{Eq}.32\right)\end{array}$

(133) Step 4. RMS-Integration.

(134) $\begin{array}{cc}{x}_{\mathrm{RMS}}\left(\mathrm{lp},t\right)=\sqrt{\frac{1}{T}\underset{i=1}{\overset{T}{.Math.}}{x_A\left(\mathrm{lp},t-i\right)}^2}& \left(\mathrm{Eq}.33\right)\end{array}$

(135) In the final steps, for example L.sub.Aeq,15(lp,t) or L.sub.C,pk(lp,t) or L.sub.EX,8h(lp,t) may be calculated, analogously to above descriptions, but with individual estimates for each listening-position lp.

(136) In other embodiments, the estimates EST1, EST2 for the different listening positions may be calculated from details of a level-calibrated system as described with reference to FIG. 4 above, an inter-channel correlation-corrected embodiment as described with reference to FIG. 6 above, or the embodiments described below.

(137) Frequency Response Characteristics

(138) FIG. 10 shows a diagram of an embodiment with an input signal IS feeding a sound system SS consisting of an amplifier connected to multiple loudspeakers which produce sound in an acoustic environment AE. Also fed by the input signal is a Mic-less sound level estimator, MLE comprising a Simulating sound level detector SSLD, capable of producing estimates EST of SPL and/or Sound Exposure, SPL_EST. The embodiment of FIG. 10 resembles that of FIG. 4 with the addition that the Mic-less sound level estimator MLE in the present embodiment further comprises frequency response characteristics FRC of loudspeakers in the acoustic environment AE and a frequency weighting filter FWF for applying the frequency response characteristics to the input signal IS in connection with the estimation of SPL or SE. The frequency weighting filter may alternatively be implemented by means of a filter bank.

(139) Thereby this embodiment is enabled to take into consideration the frequency-specific influence on the sound caused by the multiple loudspeakers, the room and furniture or other environment properties, possibly in further consideration of a specific listening position. The estimated SPL for that specific setup and listening position may thereby be more accurate, compared to the basic sound level detector and integrator SSLD, for sound system SS which have a non-flat frequency response of the loudspeaker channels in the acoustic environment AE and/or non-equal sensitivity of the loudspeaker channels in the acoustic environment AE. The frequency weighting arrangement is preferably implemented in the same DSP or CPU implementing the simulating sound level detector and integrator SSLD, and also receives the input signal IS. The frequency response characteristics FRC may preferably be stored in memory included in or connected to the DSP or CPU. The frequency response characteristics FRC may be determined as an approximation of the measured frequency response, per channel. Alternatively, the frequency response characteristics FRC may be determined based on multiple measurements. Even though such a frequency response characteristics FRC would generally be a less accurate approximation for the frequency response in a specific location, it may be simpler and more robust against small spatial variations in the listening position, which could be advantageous in some applications.

(140) In an embodiment the frequency response characteristics FRC comprises frequency response data for the loudspeakers, e.g. measured by the loudspeaker manufacturer. In another embodiment the frequency response characteristics FRC are established by performing a measurement on location of the actual setup to determine the frequency response in the actual acoustic environment.

(141) In a preferred embodiment, a measured frequency response FRC is smoothed, e.g. by a 1/12-octave resolution smoothing filter. The resulting response may then advantageously be approximated by an IIR filter, e.g. using the Yule-Walker method for recursive IIR digital filter design using a least-squares fit, or using an iterative optimizationsuch as the damped Gauss-Newton methodto minimize the difference between the actual and the desired frequency response of the IIR filter.

(142) This embodiment may be preferable over the embodiment with transfer function characteristics TFC described above with reference to FIG. 8, in case computational complexity is an issue, or in case a greater robustness against the specific listening-position would be desired.

(143) The frequency response characteristics FRC may preferably represent both differences in gain for the different channels, e.g. if the sensitivity of the loudspeakers were different, or if their placement in the acoustic environment AE caused acoustical level variation, and also the differences in the actual frequency response of the channels, i.e. the relative, effective sensitivity in different spectral regions, regardless of the absolute sensitivity.

(144) Procedure to Compute an SPL_EST Employing the Frequency Response Characteristics FRC:

(145) One embodiment of Mic-less sound level estimator MLE implementing frequency response characteristics FRC and corresponding filtering is described here, where x(c,t) refers to channel c of the input signal at time t. Further details of the variables and parameters may be found in the description of the embodiment of FIG. 4 above, mutatis mutandis.

(146) Step 1. Frequency Response Filter

(147) An IIR filter, implementing the frequency response characteristics FRC per channel, can be implemented using this linear difference equation. Let FR.sub.B and FR.sub.A denote the feedforward- and feedback-coefficients, respectively.

(148) $\begin{array}{cc}{x}_{\mathrm{FR}}\left(c,t\right)=\underset{i=0}{\overset{{\mathrm{FR}}_n}{.Math.}}x\left(c,t-i\right).Math.{\mathrm{FR}}_B\left(c,i\right)-\underset{i=1}{\overset{{\mathrm{FR}}_m}{.Math.}}{x}_{\mathrm{FR}}\left(c,t-i\right).Math.{\mathrm{FR}}_A\left(c,i\right)& \left(\mathrm{Eq}.34\right)\end{array}$

(149) In an embodiment comprising filter banks instead of a frequency response weighting filter, Eq. 34 may be substituted with a simple calculation of energy per band of the filter bank, possibly combined with the A-weighting of step 2.

(150) Steps 2-6, A-Weighting, RMS-Integration, Channel-Summation, Simulated Sound Pressure, L.sub.eq-Calculation

(151) The rest of the estimation process may be performed for example as described above with reference to FIG. 4, i.e. similar to Eqs. 1-7, by using x.sub.FR instead of x in Eq. 1 and proceed through the other steps described above. In the final steps, for example L.sub.Aeq,15(lp,t) or L.sub.C,pk(lp,t) or L.sub.Ex,8h(lp,t) may be calculated, analogously to above descriptions and possibly with above-described variations or combinations, for example regarding several listening positions as described with reference to FIG. 9 above, an inter-channel correlation-corrected embodiment as described with reference to FIG. 6 above, or the embodiments described below.

(152) Experiment

(153) FIG. 11 shows, as example, a plot of a measured frequency response of a stereo sound system (nearfield, active studio monitors) in an acoustic environment AE (a studio control room), as a smoothed magnitude response at th-octave resolution (solid line, FRC1) and 1-octave resolution (dashed line, FRC2). This frequency response characteristics FRC was produced as an approximation of the measured frequency response for the Left nearfield loudspeaker (of a stereo set) in the studio, at the primary listening position, i.e. with approximately a 30 degrees angle between the phantom center (between the stereo speakers) and the actual speaker.

(154) For the experiment, a mono test signal was generated, consisting of -octave filtered pink noise, in 9 successive octaves, 5 sec of each -octave band. FIG. 12 shows a spectrogram analysis of the test signal.

(155) FIG. 13 shows a sliding L.sub.Aeq SPL estimation computed by three different embodiments of the present invention, in the experiment with the sound system and acoustic environment described above with reference to FIG. 11 and the test signal described above with reference to FIG. 12. The three different embodiments used to estimate the SPL resulting from the test signal were the basic sound level detector and integrator SSLD described above with reference to FIG. 4, the sound level detector and integrator SSLD with transfer function characterictics TFC described above with reference to FIG. 8, and the sound level detector and integrator SSLD with frequency response characterictics FRC described above with reference to FIG. 10.

(156) The upper curve SPL1 illustrated in FIG. 13 is the result of estimating the dB SPL of the test signal according to a basic sound level detector and integrator SSLD embodiment, and reflects the A-weighting employed (FIG. 5). The middle curve SPL2 in FIG. 13 was produced by further taking into account the transfer function characteristics TFC, measured for the sound system SS (one loudspeaker) and the acoustic environment AE of the experiment. The bottom curve SPL3 in FIG. 13 illustrates the result of the experiment when employing only the overall frequency response characteristics FRC, and can be seen as an approximation of the SSLD+TFC estimate (the middle curve SPL2).

(157) In contrast to the test signal used in this experiment, a realistic signal would typically contain many frequencies across the spectrum, varying in power over time. Hence the relative effect of the transfer function characteristics TFC or the frequency response characteristics FRC on the estimate would also vary over time. In an integrated measure of SPL or Sound Exposure, the effect of the transfer function characteristics TFC and frequency response characteristics FRC would likewise be integrated. In an instantaneous or short-term measure of SPL, the effect of the transfer function characteristics TFC or frequency response characteristics FRC would depend on the characteristics of the sound system SS and acoustic environment AE, combined with the frequency content and level of the input signal, at that moment.

(158) Approximation of Sound System Properties

(159) FIG. 14 illustrates an embodiment of the invention comprising a sound system SS which receives an input signal IS and produces sound in an acoustic environment AE based on the input signal. The sound system comprises an audio amplifier AA and multiple loudspeakers LS. The sound system may optionally comprise further audio processing modules, e.g. decoder, equalizer, compressor, filter, etc. The input signal is a mono-, stereo- or multichannel analog or digital audio signal and the sound system may receive the input signal from an external source or itself integrate with a radio, media player, mixer table, or other audio source establishing the input signal.

(160) The audio amplifier AA may be any kind of audio amplifier, and the number of channels should preferably correspond to the relevant loudspeaker setup. The multiple loudspeakers are located at different positions relative to the acoustic environment, and may produce audio from different audio channels and/or from common audio channels. For example, the audio amplifier may be a stereo amplifier and the multiple loudspeakers may comprise two speakers receiving different audio channels in a stereo setup or e.g. the audio amplifier is a multi-channel receiver and multiple loudspeakers comprise 6 speakers receiving different audio channels in a surround sound setup, including a subwoofer channel. In another example the multiple loudspeakers may comprise 25 speakers all rendering the same audio channel, for example for background music in a supermarket. In another example the multiple loudspeakers may comprise 8 speakers whereof three speakers are positioned along a left side of a room and all receiving a left channel, three speakers likewise positioned along a right side of the room and all receiving a right channel, and the remaining two speakers receiving a subwoofer channel.

(161) Further, a Mic-less sound level estimator MLE is provided, which is arranged to provide, on the basis of the input signal, an estimate EST of the SPL or SE produced in at least one position in the acoustic environment. In the present embodiment, the Mic-less sound level estimator comprises a Simulating sound level detector and integrator SSLD and an approximation of sound system properties SSP to base the estimates on. In the embodiment of FIG. 14, and contrary to the embodiment of FIG. 4, the sound system SS is not necessarily level-calibrated with its loudspeakers in the acoustic environment AE. Hence, in order to obtain information about the sound system in the acoustic environment to be able to estimate an SPL or SE, one or more measurements are made by a microphone MIC in the acoustic environment to produce the approximation of sound system properties. The microphone measurements may be performed once, e.g. at the time of installation and each time major changes are made to the system or acoustic environment, or regularly, e.g. at regular intervals, at each power up, before each significant use of the system, e.g. for concerts, etc. The measurements may also be desired or even required regularly for verification of the Mic-less sound level estimator.

(162) The coupling of the microphone to the acoustic environment and the Mic-less sound level estimator indicates a temporary or periodic connection, and a possibly somewhat loose relation between the microphone and the estimator, as the microphone measurements in an embodiment may be obtained, analyzed and formalised as sound system property approximations before they are transferred to the Mic-less sound level estimator. In another embodiment the measurement microphone may be directly connected to the Mic-less sound level estimator, which may perform the analysis and formalising directly on the raw measurements.

(163) To illustrate this in further detail, FIG. 15 shows a measuring and approximation step where a test signal TS, e.g. a band-limited pink noise signal, is provided as input signal to the sound system SS and produced as sound in the acoustic environment AE. The measurement microphone MIC is installed during this step to measure the resulting test sound, and the Mic-less sound level estimator MLE, or an external device, e.g. a laptop or tablet computer, establishes an approximation of sound system properties SSP based on the measured sound, and stores this with the Mic-less sound level estimator.

(164) The approximations of sound system properties may be stored in a memory, e.g. a database, together with the DSP or CPU calculating the SPL estimate, and may be stored as any suitable abstraction level, e.g. as raw measurements, as FIR or IIR filter parameters, as a ratio between in- and out-level, possibly at different frequencies, as equivalent model parameters, etc. The approximation of system properties may preferably comprise a representation of a relation between levels of the input signal TS, IS, and the resulting Sound-Pressure Levels in the acoustic environment AE, thereby resembling the knowledge of a level-calibrated sound system, and thereby facilitating using the same procedures and equations as described above with reference to the embodiments of e.g. FIG. 4. The system properties may further relate to properties of the sound system, the acoustic environment or combinations thereof, for example a transfer function or frequency response of the loudspeakers or, preferably, the loudspeakers in the particular acoustic environment, information about the location of the loudspeakers, delays, how they interact with the room at different frequencies, room characteristics such as reverberation or diffuseness.

(165) In an embodiment, measurements are made at two or more positions, by moving the microphone and repeat the test or by installing more than one microphone. Thereby information for several listening positions are obtained, which may be applied in the estimation procedure, e.g. as described above with reference to FIG. 9.

(166) In an embodiment, the test signal TS is transmitted to only one loudspeaker at a time, to measure each combination of loudspeaker and listening position. In a more advanced embodiment, a multi-channel signal with different spectral content in each channel is used as test signal. Thereby each loudspeaker produces, simultaneously, a different sound with regard to spectral content, and if the different spectral content is carefully designed, it may be possible to distinguish each individual loudspeaker from the acoustically mixed raw measurement in the analysis. Thereby approximated system properties for all loudspeakers and one or more listening positions may be produced quickly and efficiently by one or a few averaging noise bursts from the loudspeakers. This is particularly advantageous for systems where frequent measurements are required, e.g. for regular verification of the sound level estimates. The spectral content of the test signal is frequency-domain content of the signal, i.e. the power per frequency-interval. Spectral content is considered different for two channels when the frequency spectrum of one channel, for some time period, differs significantly from the frequency spectrum of the other channel, for the same time period, and this may e.g. be obtained by using different pure tones in the different channels, e.g. by having each loudspeaker produce a number of pure tones in the low, middle and high frequency ranges, yet different and non-harmonic pure tones in each loudspeaker, or by having each loudspeaker produce a band-limited noise signal, yet with different bands or insignificant overlap of bands between the loudspeakers at the same time.

(167) Proceeding to FIG. 16 an operation step is shown, where the measurement microphone is no longer present, i.e. no feedback or measurements from the acoustic environment exists, and the system works as described above with reference to e.g. FIG. 4 or FIG. 14.

(168) Given the approximation of sound system properties SSP, the Simulating sound level detector and integrator SSLD can calculate estimates EST such as Sound Pressure Level estimates SPL-EST and Sound Exposure estimates SE-EST, in real-time and on the basis of the input signal IS.

(169) The estimation process may be performed for example as described above with reference to FIG. 4, i.e. similar to eqs. 1-7, by using the approximated system properties SSP instead of the level calibration values in Eq. 4, mutatis mutandis. Thereby, for example L.sub.Aeq,15(lp,t) or L.sub.C,pk(lp,t) or L.sub.EX,8h(lp,t) may be calculated, analogously to above descriptions and possibly with above-described variations or combinations, for example regarding several listening positions as described with reference to FIG. 9 above, an inter-channel correlation-corrected embodiment as described with reference to FIG. 6 above, embodiments further considering transfer function characteristics or frequency response characteristics as described with reference to FIGS. 8 and 10 above.

(170) Reverberation Characteristics

(171) An embodiment of the Mic-less sound level estimator MLE takes into account reverberation characteristics RC of the acoustic environment. FIG. 17 shows a diagram of such an embodiment, where an acoustic space emulator ASE essentially implements a simplistic artificial reverberation that then applies reverb to the input signal IS prior to its processing by the sound level detector and integrator SSLD. In contrast to common artificial reverberation methods, the artificial reverberation added by the acoustic space emulator ASE is not meant to sound like the real acoustic environment AE at all, but only emulate those properties relevant to the subsequent estimate of SPL or SE, i.e. those properties provided in the reverberation characteristics RC. The reverberation characteristics may for example simply be the reverberation time, such as an R.sub.T60 and an Early Decay Time. Such parameters may be measured of a real room or hall (acoustic environment AE) e.g. using the Schroeder method, or they may be provided as parameters, e.g. based on knowledge of the inner dimensions of the acoustic environment AE. The reverberation characteristics RC parameters may advantageously be provided per octave, for a more detailed emulation of the acoustic space emulator ASE.

(172) FIG. 18 shows a diagram of another embodiment taking into account reverberation of the acoustic environment AE. In this embodiment the sound level detector and integrator SSLD is given auxiliary information by an acoustic space SPL emulator ASSE, by means of which it may provide a more accurate estimate of SPL or Sound Exposure. The acoustic space SPL emulator ASSE employs basic assumptions of the acoustics of the acoustic environment AE, such as its reverberation decaying at an exponential rate; the actual rate supplied in the reverberation characteristics RC. When the sound level detector and integrator SSLD calculates the short-term RMS (e.g. Eq. 14), it may thenas an additional stepfilter the x.sub.RMS to apply the extra exponential decay determined by the acoustic space SPL emulator ASSE. Thus, the effect of the reverberation on the SPL estimate is applied in the RMS-domain instead of applying the reverberation to the actual input signal IS (which would be computationally more complex).

(173) The embodiments of FIG. 17 and FIG. 18 with reverberation characteristics RC and acoustic space emulation ASE or acoustic space SPL emulation ASSE may be combined with any of the other embodiments described above, e.g. by considering further listening positions, inter-channel correlation, transfer function or frequency response characteristics of the loudspeakers, various calibration methods and test signals, etc.

(174) Auxiliary Microphone

(175) In variations of any of the embodiments described above, e.g. with reference to FIG. 4, FIG. 6, FIG. 8, FIG. 9, FIG. 10, FIG. 14, FIG. 15, FIG. 16, FIG. 17 or FIG. 18, an auxiliary microphone may be added to the system. In preferred embodiments, the auxiliary microphone is not for measuring sound level at a listening position, as the Mic-less sound level estimator MLE of the present invention is also estimating, e.g., SPL or SE, in embodiments with the auxiliary microphone. However, in preferred embodiments, the auxiliary microphone may for example be permanently installed at a location within the acoustic environment where it may remain visually discrete and non-disturbing, and/or not be in the way of people moving around, rearranging furniture, etc. In embodiments with an auxiliary microphone, it may be used to provide feedback to the Mic-less sound level estimator MLE about the sound present at the location of the auxiliary microphone. In several setups the feedback from the auxiliary microphone is not usable directly for comparing with the estimated SPL, but the feedback from the auxiliary microphone may preferably be analyzed to detect indication of assumptions and approximations of the loudspeakers, acoustic environment, etc., being incorrect or changed since last calibration. In an embodiment, the sound measured by the auxiliary microphone in response to a test signal, e.g. during calibration, is stored, or properties thereof are stored, e.g. as a parametrized version of the sound or as key properties such as amplitude, phase, spectrum, etc. During normal use or later calibrations, sound measured by the auxiliary microphone may be compared with the stored version, properties or parameters, thereby enabling a detection of changes or errors. In an embodiment, the comparing of measurements from the auxiliary microphone with the stored reference measurement, properties or parameters, is performed continuously or frequently and the result used to make slight adjustments to the sound level estimates, or to detect errors. Thereby slowly changing properties that may assumedly be less location-specific, e.g. physical wear or deformation in the loudspeakers, changes of temperature or static air pressure, etc., may be compensated for in the estimate without requiring a new calibration. Emerging errors which affect the generated sound, e.g. a disconnection of a loudspeaker, a damaged loudspeaker, bad cable connections, etc., may also be detected in an embodiment with an auxiliary microphone. The analysis of the signal from the auxiliary microphone may be considering the sound system as a whole, or consider each channel individually, e.g. to determine a loudspeaker disconnect. In various embodiments, the auxiliary microphone needs not be a high-quality or super-sensitive microphone, as the sound measured thereby is not necessarily used as an audio signal, but preferably rather properties thereof. The auxiliary microphone may be any kind of transducer responsive to acoustic sound. In an embodiment several auxiliary microphones are provided. At a block-diagram level, an embodiment comprising an auxiliary microphone may be similar to e.g. one of the embodiments shown in FIG. 14 or 15, by using the microphone MIC as auxiliary microphone instead of measurement microphone.