Method for transmitting data representing ultrasonic measurement signals, in particular in a vehicle

Abstract

In the method for transmitting data representing an ultrasonic measurement signal of an ultrasonic measuring device, in particular for a vehicle, from a transmitter to a receiver a digitized analog ultrasonic measurement signal is provided in the transmitter. On the transmitter side the ultrasonic measurement signal is sampled at a multiple of its frequency and divided into individual successive blocks of sampling values. The sampling values of the sampled ultrasonic measurement signal are transformed in blocks into the frequency range. Those frequency portions of the spectrum whose amplitude is smaller than a presettable threshold value, or the frequency portions of the spectrum above an upper frequency limit value and/or below a lower frequency limit value are removed. The amplitude range covered by the remaining frequency spectrum is scaled by a scaling factor for further reduction of the data. The data of each block with the scaling factor assigned to the respective block are transmitted to the receiver. On the receiver side the scaling of the amplitude range of the frequency spectrum of each block is reversed using the respective scaling factor and the frequency spectrum is transformed back into the time range.

Claims

1. A method for transmitting data representing an ultrasonic measurement signal of an ultrasonic measuring device, in particular for a vehicle, from a transmitter to a receiver, wherein in the method: in the transmitter a digitized analog ultrasonic measurement signal is provided in reaction to an analog ultrasonic transmission signal emitted for detecting obstacles, the ultrasonic measurement signal is sampled at a multiple of its frequency and divided into individual successive blocks of sampling values, the sampling values of the sampled ultrasonic measurement signal are transformed in blocks into the frequency range using a segmented fast convolution, those frequency portions of the spectrum whose amplitude is smaller than a presettable threshold value, or the frequency portions of the spectrum above an upper frequency limit value and/or below a lower frequency limit value are removed, the amplitude range covered by the remaining frequency spectrum is scaled by a scaling factor for further reduction of the data, and from the transmitter the data of each block with the scaling factor assigned to the respective block are transmitted to the receiver, and in the receiver the scaling of the amplitude range of the frequency spectrum of each block is reversed using the respective scaling factor, the thus processed frequency spectrum is filtered out of the analog ultrasonic measurement signal provided in the transmitter by multiplication by filter coefficients of an optimum filter for extracting the signal shape of the analog ultrasonic transmission signal, and the thus filtered frequency spectrum is transformed in blocks back into the time range using an inverse segmented fast convolution.

2. The method according to claim 1, wherein the scaling is carried out by identifying, in an amplitude bit word having L bits and describing an amplitude value, that most significant bit of these L bits which is not equal to zero, and this most significant bit as well as, starting therefrom, the M next less significant bits of the amplitude bit word are converted to a reduction bit word of the length M+1 with M+1<L, wherein the scaling factor represents the number of those bits of the amplitude bit word which are more significant than the identified most significant bit not equal to zero of the amplitude bit word.

3. The method according to claim 2, wherein, when the identified most significant bit not equal to zero of the amplitude bit word is less significant than the (M+1)th bit, calculated as from the least significant bit of the amplitude bit word, these (M+1) bits of the amplitude bit words are the bits of the reduction bit word and the scaling factor represents the number L(M+1).

4. The method according to claim 1, wherein the digitized analog ultrasonic measurement signal is subjected to an I/Q demodulation, wherein the frequency at which the I/Q demodulation is carried out is equal to the frequency of the ultrasonic measurement signal.

5. The method according to claim 2, wherein the digitized analog ultrasonic measurement signal is subjected to an I/Q demodulation, wherein the frequency at which the I/Q demodulation is carried out is equal to the frequency of the ultrasonic measurement signal.

6. The method according to claim 3, wherein the digitized analog ultrasonic measurement signal is subjected to an I/Q demodulation, wherein the frequency at which the I/Q demodulation is carried out is equal to the frequency of the ultrasonic measurement signal.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

(1) Hereunder the invention is explained in detail on the basis of an exemplary embodiment with reference to the drawings in which:

(2) FIG. 1 shows the principle according to the invention of the USPA data compression,

(3) FIG. 2 shows a schematic representation for illustrating the fast convolution,

(4) FIG. 3 shows a schematic representation for illustrating the segmented convolution,

(5) FIG. 4 exemplarily shows a spectrum of a 16-pulse ultrasonic burst in the sensor in relation to the center frequency,

(6) FIG. 5 shows a schematic representation of the scaling,

(7) FIG. 6 shows a flow chart for an exemplary embodiment of the USAP data compression, and

(8) FIGS. 7 and 8 show curve shapes for comparison before and after a data compression.

DETAILED DESCRIPTION OF A PREFERRED EMBODIMENT OF THE INVENTION

(9) In the conventional ultrasonic parking assistance systems, the signal processing of the receive channel is fully integrated in each sensor module. Besides an amplifier and an ADC, it is composed of the following digital function blocks: demodulation/quadrature mixer matched filter/pulse compression filter/correlation meter absolute-value calculator algorithms for rendering more precise the echo signals and for interference suppression, such as differentiation (FTC=fast time constant) algorithms for automatic and manual threshold value generation determination of the echo position (e.g. echo peak detection) echo evaluation according to height/surface area/signal-to-noise ratio etc.

(10) These functions are almost exclusively implemented in digital hardware. This is cost-effective but also inflexible. Although this drawback can be compensated for by an extensive parametrization, the actual algorithms can no longer be modified. An implementation in software on a correspondingly powerful digital signal processor (DSP) in each sensor module would be the solution but is too expensive at least for the time being.

(11) When the echo data are combined in the central control unit, normally only the echo position and in very few cases also the echo height are available. For a more precise recognition of the environment it is desired to provide the complete envelope having a high resolution to the control unit. If the main part of the signal processing were realized in the central unit in the form of software, a very flexible system having functionally simple and standardized sensor modules would be available.

(12) In the case of currently used sensors the envelope is available in a time grid of 20 s with a 16-bit resolution. This means that per measuring period of 5 msec 500 bytes of data would have to be transmitted, for example. In contrast, the currently provided DSI3 bus which is considered fast has a net transmission rate of merely approximately 150 bytes/5 msec. It is therefore an object to provide a method where the amount of data is reduced without significantly affecting the signal quality. Such a system is schematically shown in FIG. 1.

(13) 1. Solution

(14) The mathematical basis of the solution is the discrete convolution which is roughly described here. Literature concerning the discrete convolution is stated under [1], [2], [3], for example.

(15) 1.1 Fast Convolution

(16) The filtering of a signal can be carried out by convolution with the pulse response of the filter:

(17) $g\left(t\right)=s\left(t\right)*h\left(t\right)={}_{-}^{}s\left(\right)h\left(t-\right)d$

(18) For discrete sequences the following can be derived:

(19) $g\left(n\right)=s\left(n\right)*h\left(n\right)=\underset{m=-}{\overset{}{.Math.}}s\left(m\right)h\left(n-m\right)$

(20) The length of the convolution product is
L=M.sub.s+M.sub.h1

(21) For the Fourier transform of the discrete convolution the following applies:
s(n)*h(n)S(f).Math.H(f)

(22) i. e. after the transformation the convolution operation proceeds to a multiplication. However, depending on the length of the sequences, this apparent detour allows a lot of operations to be omitted, which explains the term fast convolution.

(23) The calculation of the fast convolution using the fast Fourier transform (FFT) proceeds as follows (see also the schematic representation in FIG. 2): 1. Filling the sequences with zeros up to the length L 2. Fourier transformation of the sequences 3. Multiplication of the transformed sequences 4. Back-transformation (inverse Fourier transformation)

(24) 1.2 Segmented Convolution

(25) In many cases the signal sequence is very long and/or infinitely long (continuous signal processing). Frequently, the memory is not large enough or no more delay/latency can be accepted. In this case the input sequence is partitioned into subsequences. After the convolution these sequences are added as is shown in the following figure (overlap-and-add-methodsee FIG. 3), for example.

(26) 1.3 Description of the Invention on the Basis of an Exemplary Embodiment

(27) The calculation of the envelope and the processing of the following algorithms are carried out in the time grid of 20 s (sampling frequency 50 kHz). The resolution is necessary for actually attaining the required absolute accuracy at the distance of <=1 cm (20 s correspond to 3.4 mm).

(28) The frequency spectrum of an echo signal in the receiver, which has been emitted with 16 pulses, is observed. As compared to the representable bandwidth of +/25 kHz the echo signal comprises a significant bandwidth of a maximum of +/3 kHz (see FIG. 4).

(29) Thus, if the signal transmission is carried out in the frequency range instead of the time range the major portion of the signal is obviously negligible. In this connection it is useful to place the matched filter also into the central control unit, in particular if it is a complex pulse compression filter (e.g. for chirp signals). Since the echo data are available as frequency data in the central control unit they can be very efficiently filtered in the central control unit according to the principle of a fast convolution. For reducing the storing operation in the sensor and above all the latency time the measurement data are transformed in blocks into the frequency range. The basis for this is the so-called segmented convolution. In addition, the segmentation offers the advantage that each section can be separately scaled and thus the amount of data can be reduced again at least by a factor of two. During the transmission this factor is added to the respective data package to reverse the scaling in the central unit.

(30) An example of the scaling of a bit word of the length L (L=16 bits) to a bit word of the length M+1 (e.g. 8 bits) is shown in FIG. 5. The most significant bit, which is not equal to zero, is searched in the bit word of the length L. In the example of FIG. 5 this is the fourth-most-significant bit. This bit as well as the next seven less significant bits are used as a reduced bit word of the length 8. The scaling factor is 3.

(31) If the uppermost 8 bits of the exemplary 16-bit word are zero, the reduced bit word of the length 8 comprises the lowermost 8 bits of the 16-bit word.

(32) The application flow of the data compression, the transmission of the data as well as the back-transformation are shown in FIG. 6.

(33) The signal is received by the transducer (receiver) and subsequently amplified and digitized. With the aid of a quadrature demodulator it is shifted to the frequency zero and partitioned into the in-phase portions (I) and the quadrature portions (Q). The sinc filter eliminates the higher-frequency components, in particular the portions are precisely eliminated at the mixer frequency (mainly caused by the offset of the amplifier and the ADC) and at twice the mixer frequency. Subsequently, the sampling rate is reduced, e. g. to one value per signal period (corresponding to 20 s, see above). The thus produced signal is also referred to as an analytic signal with I as a real portion and Q as an imaginary portion [4]. The complex-valued signal is now transformed in blocks into the frequency range using the fast Fourier transformation (FFT). Those frequency portions which exceed a specific limit frequency fg are rejected. The remaining values are now scaled (e. g. using a shift operation, multiplication by 2) such that they can be represented by a considerably shorter data word (e. g. 8 instead of 16 bits). Thereafter the transmission via the data bus is carried out.

(34) In the central unit the data are first rescaled to their original value range. Subsequently, the filtering in the frequency range by multiplication by the frequency response of the filter is carried out. This is advantageous in that the filter need not be calculated: the frequency response is simply saved in the memory. In the last step a transformation back into the time range is carried out using an inverse Fourier transformation (IFFT). The further processing of the echo data, such as evaluation with static or dynamic threshold values (CFAR) etc. (see above), is subsequently carried out in a known manner.

(35) The curves of FIGS. 7 and 8 show the differences between a data-reduced and a not-data-reduced signal. Transmission frequency 50 kHz, sampling time after demodulation 20 s Original data rate: 500 bytes/5 msec Reduced data rate: 62 bytes/5 msec Bandwidth (as per FFT bins): +/2.4 kHz FFT length L=64 Length of pulse response of the matched filter Nh=16 Block size: Ns=49 sampling values corresponding to 0.98 msec Number of FFT bins taken into account (frequency values): 7 Word width of the transmitted values: 6 bits

(36) Although the invention has been described and illustrated with reference to a specific illustrative embodiment thereof, it is not intended that the invention be limited to that illustrative embodiment. Those skilled in the art will recognize that variations and modifications can be made without departing from the true scope of the invention as defined by the claims that follow. It is therefore intended to include within the invention all such variations and modifications as fall within the scope of the appended claims and equivalents thereof.

LITERATURE

(37) [1] Albrecht Ludloff, Praxiswissen Radar and Radarsignalverarbeitung [2] https://de.wikipedia.org/wiki/Schnelle_Faltung [3] U.S. Pat. No. 5,502,747 [4] https://de.wikipedia.org/wiki/Analytisches_Signal