Method for Distance Determination

Abstract

Example embodiments relate to methods for distance determination. One embodiment includes a method for determining a distance between a first radio signal transceiver and a second radio signal transceiver. The method includes receiving a first set and a second set of measurement results. The first set is acquired by the first radio signal transceiver based on signals transmitted from the second radio signal transceiver and the second set is acquired by the second radio signal transceiver based on signals transmitted from the first radio signal transceiver. The first set is representable as including, for each of a plurality of frequencies, a measurement pair. The method also includes calculating, for each frequency of the plurality of frequencies, a preliminary estimate of a value proportional to a one-way frequency domain channel response. Additionally, the method includes determining a final estimate of the value proportional to the one-way frequency domain channel response.

Claims

1. A method for determining a distance between a first radio signal transceiver and a second radio signal transceiver, wherein the method comprises: receiving a first set of measurement results and a second set of measurement results, wherein the first set of measurement results is acquired by the first radio signal transceiver based on signals transmitted from the second radio signal transceiver and the second set of measurement results is acquired by the second radio signal transceiver based on signals transmitted from the first radio signal transceiver, wherein the first set of measurement results is representable as comprising, for each of a plurality of frequencies, a measurement pair of a phase value and a signal strength value and the second set of measurement results is representable as comprising, for each of the plurality of frequencies, a phase value or a measurement pair of a phase value and a signal strength value; calculating, for each frequency of the plurality of frequencies, a preliminary estimate of a value proportional to a one-way frequency domain channel response, wherein the preliminary estimate is based on: the measurement results for the frequency from the first set of measurement results and the phase value; or the measurement results for the frequency from the second set of measurement results; calculating, for a frequency of the plurality of frequencies, a predicted estimate of a representation of the value proportional to the one-way frequency domain channel response, wherein the predicted estimate is based on one or more estimates for one or more respective frequencies adjacent to the frequency; calculating a first metric distance between the predicted estimate for the frequency and a representation of the preliminary estimate for the frequency; calculating a second metric distance between the predicted estimate for the frequency and a phase reversal of the representation of the preliminary estimate for the frequency; determining, for the frequency, a final estimate of the value proportional to the one-way frequency domain channel response, wherein the final estimate is based on a comparison of the first metric distance and the second metric distance, and wherein the final estimate is either a phase reversal of the preliminary estimate or the preliminary estimate; and determining the distance between the first radio signal transceiver and the second radio signal transceiver based on a plurality of such final estimates.

2. The method of claim 1, wherein the predicted estimate is a representation of an estimate for an immediately adjacent frequency.

3. The method of claim 1, wherein the predicted estimate is an extrapolation from two or more estimates for two or more respective adjacent frequencies to the frequency.

4. The method of claim 3, wherein the two or more respective adjacent frequencies the frequency are lower than the frequency, and wherein the method further comprises: calculating a second predicted estimate of a representation of a value proportional to a one-way frequency domain channel response, wherein the second predicted estimate is based on one or more estimates for the frequency and one or more respective frequencies adjacent to and higher than the frequency; calculating a third metric distance between the second predicted estimate and a representation of a preliminary estimate corresponding to the second predicted estimate; and calculating a fourth metric distance between the second predicted estimate and a phase reversal of the representation of the preliminary estimate corresponding to the second predicted estimate, wherein the final estimate of the value proportional to the one-way frequency domain channel response is based on comparison of the first metric distance, the second metric distance, the third metric distance, or the fourth metric distance.

5. The method of claim 1, wherein each representation of a respective quantity is the quantity itself.

6. The method of claim 1, wherein each representation of a respective quantity is a frequency-dependent transformation compensating for an inherent phase advance between adjacent frequencies.

7. The method of claim 6, wherein the frequency-dependent transformation is calculated as an average phase advance between the plurality of frequencies.

8. The method according to claim 1, wherein calculating the preliminary estimate of the value proportional to the one-way frequency domain channel response comprises: calculating an estimate of a value proportional to a two-way frequency domain channel response based on the measurements for the frequency from the first set of measurement results and the phase value or the measurements for the frequency from the second set of measurement results; and calculating the preliminary estimate proportional to the one-way frequency domain channel response based on the estimate of the value proportional to the two-way frequency domain channel response.

9. The method according to claim 1, wherein each measurement pair of the measurement pairs is representable as a first complex number, wherein a modulus of the first complex number represents an amplitude corresponding to the signal strength value and an argument the first complex number represents the phase value, wherein the preliminary estimate and the final estimate each are representable by second complex numbers, and wherein a modulus of the second complex number represents an amplitude response and the argument of the second complex number represents a phase response.

10. The method according to claim 9, wherein the second set of measurement results comprises, for each of the plurality of frequencies, the measurement pair, and wherein calculating the estimate of the value proportional to a two-way frequency domain channel response comprises, or is representable as comprising, multiplying the first complex number representing the measurement pair from the first set of measurement results with the first complex number representing the measurement pair from the second set of measurement results.

11. The method according to claim 9, wherein calculating the preliminary estimate of the value proportional to the one-way frequency domain channel response based on the estimate of the value proportional to the two-way frequency domain channel response comprises, or is representable as comprising, taking a complex square root of the estimate proportional to the two-way frequency domain channel response.

12. The method according to claim 11, wherein, when taking the square root, a solution with a phase between −π/2 and π/2 is selected.

13. The method according to claim 1, wherein determining the distance between the first radio signal transceiver and the second radio signal transceiver uses an algorithm based on IFFT or a super-resolution algorithm.

14. A computer program product comprising a computer-readable medium storing computer-readable instructions such that, when executed on a processing unit, the computer program product will cause the processing unit to perform the method according to claim 1.

15. A radio signal transceiver configured to determine a distance to a second radio signal transceiver, wherein the radio signal transceiver comprises: a measurement unit configured to acquire a first set of measurement results based on signals transmitted from the second radio signal transceiver, wherein the first set of measurement results is representable as comprising, for each of a plurality of frequencies, a measurement pair of a phase value and a signal strength value; a receiver configured to receive a second set of measurement results acquired by the second radio signal transceiver based on signals transmitted from the first radio signal transceiver, wherein the second set of measurement results is representable as comprising, for each of a plurality of frequencies, a phase value or a measurement pair of a phase value and a signal strength value; and a processing unit configured to: calculate for each frequency of the plurality of frequencies, a preliminary estimate of a value proportional to a one-way frequency domain channel response, wherein the preliminary estimate is based on: the measurement results for the frequency from the first set of measurement results and the phase value; or the measurement results for the frequency from the second set of measurement results; calculate, for a frequency of the plurality of frequencies, a predicted estimate of a representation of the value proportional to the one-way frequency domain channel response, wherein the predicted estimate is based on one or more estimates for one or more respective frequencies adjacent to the frequency; calculate a first metric distance between the predicted estimate for the frequency and a representation of the preliminary estimate for the frequency; calculate a second metric distance between the predicted estimate for the frequency and a phase reversal of the representation of the preliminary estimate for the frequency; determine, for the frequency, a final estimate of the value proportional to the one-way frequency domain channel response, wherein the final estimate is based on a comparison of the first metric distance and the second metric distance, and wherein the final estimate is either a phase reversal of the preliminary estimate or the preliminary estimate; and determine the distance between the first radio signal transceiver and the second radio signal transceiver based on a plurality of such final estimates.

16. The radio signal transceiver of claim 15, wherein the predicted estimate is a representation of an estimate for an immediately adjacent frequency.

17. The radio signal transceiver of claim 15, wherein the predicted estimate is an extrapolation from two or more estimates for two or more respective adjacent frequencies to the frequency.

18. The radio signal transceiver of claim 15, wherein each representation of a respective quantity is the quantity itself.

19. The radio signal transceiver of claim 15, wherein each representation of a respective quantity is a frequency-dependent transformation compensating for an inherent please advance between adjacent frequencies.

20. The radio signal transceiver of claim 19, wherein the frequency-dependent transformation is calculated as an average phase advance between the plurality of frequencies.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

[0043] The above, as well as additional objects, features and advantages of the present inventive concept, will be better understood through the following illustrative and non-limiting detailed description, with reference to the appended drawings. In the drawings like reference numerals will be used for like elements unless stated otherwise.

[0044] FIG. 1 is a timing diagram of transmissions between two transceivers.

[0045] FIG. 2 is a flowchart summarizing steps of the method.

[0046] FIG. 3a shows a simulated example of a preliminary estimate of a one-way frequency-domain channel response.

[0047] FIG. 3b shows a final estimation of the one-way frequency-domain channel response of FIG. 3a.

[0048] FIG. 3c shows the preliminary estimate of FIG. 3a as transformed to compensate for an inherent phase advance between adjacent frequencies.

[0049] FIG. 3d shows a final estimation derived from the estimate of FIG. 3c.

[0050] FIG. 4 illustrates extrapolation.

[0051] FIG. 5 shows a simulation geometry.

[0052] FIG. 6a, 6b, 6c, 6d show simulated reconstruction performance in the form of channel quality indicator.

DETAILED DESCRIPTION

[0053] The disclosed method may, using the multicarrier phase difference (MCPD) ranging principle, for a range determination, i.e., a distance determination, between a first radio signal transceiver, device A, and a second radio signal transceiver, device B, use as input a first set of measurement results and a second set of measurement results, wherein the first set of measurement results is acquired by the first radio signal transceiver, i.e., device A, based on signals transmitted from the second radio signal transceiver, i.e., device B, and the second set of measurement results is acquired by the second radio signal transceiver, i.e., device B, based on signals transmitted from the first radio signal transceiver. Each set of measurement results comprises, for each of a plurality of frequencies, a measurement pair of a phase measurement and a signal strength measurement.

[0054] Acquiring of the measurement results may start with the two devices A and B agreeing on the ranging parameters, align their frequencies (e.g. using carrier frequency offset (CFO) estimation and calibration) and realize coarse time synchronization, i.e. both A and B start a (digital) counter, i.e, clock at, e.g., the transmission/reception of a start frame delimiter (SFD) which both devices A and B use to control a local state machine. The state machine controls when which transceiver is doing what.

[0055] As illustrated in FIG. 1, the measurements may be performed in the following steps: [0056] 0. Device A and Device B setting their respective local oscillators [0057] (LOs) to a predetermined frequency and setting the loop counter k=0. [0058] 1. Device A transmitting its LO signal and Device B performing a phase measurement ϕ.sub.B[k]) relative to its own LO. Further, device B may perform a received signal strength indication measurement RSSI.sub.B[k] (not shown). Alternatively, device B may perform a measurement of the cartesian I.sub.B[k] (in-phase) and Q.sub.B[k] (quadrature) components of the signal received at B, relative to its own LO. [0059] 2. Device A and B changing transmit direction, allowing a guard time for stabilizing the LO. [0060] 3. Device B transmitting its LO signal and Device A performing a phase measurement ϕ.sub.A[k]) relative to its own LO. Further, device A may perform a received signal strength indication measurement RSSI.sub.A[k] (not shown). Alternatively, device A may perform a measurement of the cartesian I.sub.A[k] (in-phase) and Q.sub.A[k] (quadrature) components of the signal received at A, relative to its own LO. [0061] 4. Device A and Device B increasing the frequency of their respective LOs by a predetermined frequency spacing Δ.sub.f and go back to step 1. This loop is repeated a predetermined number of times (K.sub.f), resulting in measurements at K.sub.f different frequencies with a spacing Δ.sub.f and ordered in frequency according to their respective frequency index k. For example, measurements may be performed with a 1 MHz frequency spacing of an 80 MHz band at 2.4 GHz.

[0062] Device A and device B have respective phase-locked loops (PLLs) to generate their respective local oscillator (LO) signals. When switching from transmit to receive or vice-versa, for each single frequency k, the PLLs remain on to allow for continuous phase signals.

[0063] FIG. 2 summarizes steps of the method. In block 2 of FIG. 2, once the measurements have been carried out, the method for determining the distance is not very time-critical. Therefore, it may be computed on a third device/entity with more processing power, which is, e.g., in the cloud, assuming the entity has access to the measurement data from both transceivers. Thus, the method may either be performed on device A and/or B, but may also be collected on a third Device C, which can then calculate the distance between A and B, where device C may be in the cloud. If a device is not to carry out the method, it may transmit, or cause to be transmitted its measurement results to the device that is to carry out the method. Thus, for example, if the method is to be carried out on device C, device B may transmit a frame with all its phase measurements to Device C (ϕ.sub.B[0:K.sub.f−1]) and device A may transmit a frame with all its phase measurements to Device C (ϕ.sub.A[0:K.sub.f−1]). Similarly, the RSSI measurements RSSI.sub.A[0:K.sub.f−1] and, optionally, RSSI.sub.B[0:K.sub.f−1] will be transmitted to the device carrying out the method, for example device C. Thus, the device carrying out the method receives the first set of measurement results, comprising OA[0:K.sub.f−1] and RSSI.sub.A[0:K.sub.f−1] (or I.sub.A[0:K.sub.f−1] and Q.sub.A[0:K.sub.f−1]) and the second set of measurement results, comprising OB[0:K.sub.f−1] and, optionally, RSSI.sub.B[0:K.sub.f−1] (or I.sub.B[0:K.sub.f−1] and Q.sub.B[0:K.sub.f−1].

[0064] Alternatively, device A will carry out the method and may then comprise a measurement unit configured to acquire the first set of measurement results based on signals transmitted from the second radio signal transceiver, i.e., device B, as per the above. It may further comprise a receiver configured to receive the second set of measurement results acquired by the second radio signal transceiver, i.e., device B, based on signals transmitted from the first radio signal transceiver, i.e., device A. Further, device A may comprise a processing unit for carrying out the steps of the method, as will be described below.

[0065] For each frequency and set of measurements, a complex number may be formed, proportional to the one-way frequency domain response, where the modulus represents an amplitude corresponding to the signal strength measurement and the argument of said complex number represents the phase measurement:

H.sub.A[k]=A.sub.A[k]exp(jϕ.sub.A[k])

H.sub.B[k]=A.sub.B[k]exp(jϕ.sub.B[k])

where A.sub.A[k] and A.sub.B [k] are values proportional to signal amplitude, obtainable, for example, by taking the square root of the corresponding RSSI values.

[0066] Alternatively, in the case of measurement of the I and Q components of the signal, H.sub.A [k] and H.sub.B [k] may be formed thus:

H.sub.A[k]=I.sub.A[k]+j Q.sub.A[k]

H.sub.B[k]=I.sub.B[k]+j Q.sub.B[k]

[0067] In the absence of thermal or phase-noise, these measured magnitudes and phases at the kth frequency are related to the actual channel responses H [k] as follows

H.sub.A[k]∝H[k]exp(j2πθ[k])

[0068] where θ[k] denotes a phase offset between A and B during the measurement of the kth frequency and the symbol a denotes proportionality, i.e., a[k]∝b[k] means that a[k]=c b[k] for all values of k, where c is an unknown complex-value, but the same for all k.

[0069] Contrary to the method disclosed in EP3502736A1, there is no restriction on θ[k], which can be allowed to vary arbitrarily from frequency to frequency k.

[0070] Similarly, at B, we will measure

H.sub.B[k]∝H[k]exp(−j2πθ[k])

[0071] In block 4 of FIG. 2, an estimate X[k] of a value proportional to a two-way frequency domain channel response can be formed by multiplying the two values together, thereby canceling out the factors related to the offsets θ[k], since, as the measurements at A and B are taken shortly after each other, we may assume that θ[k] did not change:

X[k]=H.sub.A[k]H.sub.B[k]∝(H[k]).sup.2.

[0072] Thus, the calculation of the estimate of a value proportional to the two-way frequency domain channel response is based on the measurement pair from the first set of measurement results and the measurement pair from the second set of measurement results. Moreover, it comprises, or may be represented as comprising, multiplying the complex number representing the measurement pair from the first set of measurement results with the complex number representing the measurement pair from the second set of measurement results.

[0073] Alternatively, X[k] may, regarding amplitude, be calculated based on the measurement at A only:

X[k]=∥H.sub.A[k]∥.sup.2 exp(ϕ.sub.A[k]+ϕ.sub.B[k])∝(H[k]).sup.2

where ∥ ∥.sup.2 denotes the absolute squared-operator. Note that ∥ H.sub.A[k]∥.sup.2 is equal to the RSSI.sub.A[k].

[0074] Thus, here, calculating the estimate of a value proportional to the two-way frequency domain channel response is based on the measurement pair from the first set of measurement results and the phase measurement from the second set of measurement results.

[0075] In the following, the one-way frequency-domain channel response H[k] will be reconstructed using a) X[k] and b) correlation properties for H[k] for adjacent frequencies.

[0076] A preliminary estimate H.sub.sqrt [k] of the one-way frequency domain channel response H[k] is calculated by taking the square root of the estimate proportional to the two-way frequency domain channel response X[k]:

H.sub.prelim[k]=H.sub.sqrt[k]=√{square root over (X[k])}∝c[k]H[k]

[0077] which is related to the true one-way frequency-domain channel response according to the proportionality above, where c[k] is either +1 or −1, caused by the inherent phase ambiguity of taking a complex square root. To estimate the values of c[k], we use the correlation properties for H[k].

[0078] Thus, starting from the estimated frequency-domain channel response H.sub.prelim[k] that contains random phase reversals, i.e., sign flips at various frequency indices k, we want to detect those sign-flips (or, equivalently, the signs in c) and corrects them to restore the phase structure along the frequency dimension.

[0079] For the preliminary estimate, for example, solutions with the phase between π/2 and π/2 may be selected, i.e., with a positive real part.

[0080] Thus, for each frequency, the preliminary estimate of the value proportional to the one-way frequency domain channel response is calculated based on the measurement pair from the first set of measurement results and the phase measurement, or optionally the measurement pair, from the second set of measurement results.

[0081] At block 5a, optionally, each preliminary estimate H.sub.prelim[k] may undergo a transformation, which may be frequency dependent, resulting in a representation H′.sub.prelim[k] of each preliminary estimate, as will be exemplified further below.

[0082] Alternatively, as exemplified immediately below, calculations may be performed directly on values of H.sub.prelim[k]. In that case, each said representation H′.sub.prelim[k] of a quantity is the quantity itself, i.e., H′.sub.prelim[k]=H.sub.prelim[k].

[0083] At block 5b, for a frequency index k in the plurality of frequencies, a predicted estimate H′.sub.pred [k] of H′[k] is calculated based values of H′.sub.prelim[k] for frequency indices adjacent to k, i.e., H′.sub.prelim[k−1], H′.sub.prelim[k+2] . . . and/or H′.sub.Prelim [k+1], H′.sub.prelim[k+2] . . . .

[0084] In the simplest case, the predicted estimate H′.sub.pred [k] of H′[k] may be calculated as the value of the estimate for the preceding frequency index k−1

H′.sub.pred[k]=H′.sub.prelim[k−1].

[0085] Alternatively, the predicted estimate may be based on extrapolation from two or more adjacent frequency indices, as will be exemplified below.

[0086] At block 6, a first metric distance d.sub.L between the predicted estimate H′.sub.pred [k] and the preliminary estimate H′.sub.prelim[k] is calculated:

d′.sub.L=∥H′.sub.prelim[k]−H′.sub.pred[k]∥.

[0087] Still at block 6, a second metric distance d′.sub.L between the predicted estimate H′.sub.pred [k] and a phase reversal—H′.sub.prelim[k] of the preliminary estimate H′.sub.prelim[k] may be calculated as:

d′.sub.L=∥−H′.sub.prelim[k]−H′.sub.pred[k]∥=∥H′.sub.prelim[k]+H′.sub.pred[k]∥.

[0088] The metric according to which the first metric distance d.sub.L and the second metric distance d′.sub.L is calculated may be the complex number norm, or some other geometric metric, i.e., distance measure.

[0089] At block 8, a final estimate H.sub.est[k] of the one-way frequency domain channel response is determined. The final estimate H.sub.est[k] is based on a comparison of the predicted estimate H′.sub.pred [k] with the preliminary estimate H.sub.Pred[k] and its phase reversal—H′.sub.pred [k], more specifically based on a comparison of the first metric distance d.sub.L and the second metric distance d′.sub.L, as calculated at block 6.

[0090] Under an assumption of Gaussian-distributed random errors, d.sub.L is a measure for the likelihood given no sign-flip, i.e., no phase reversal is required in the final estimation relative to the preliminary estimation and d′.sub.L is a measure for the likelihood that a phase reversal is required.

[0091] Thus, if d′.sub.L<d.sub.L, this indicates an incorrect sign in the preliminary estimate H.sub.prelim[k] and the final estimate is determined to be

H.sub.est[k]=−H.sub.prelim[k].

[0092] Otherwise, the final estimate is determined to be

H.sub.est[k]=H.sub.prelim[k].

Thus, the final estimate H.sub.est[k] is either a phase reversal—H.sub.prelim[k] of the preliminary estimate, or the preliminary estimate H.sub.prelim[k].

[0093] Further, the representation of the preliminary estimate may be updated

H′.sub.prelim[k]=H′.sub.est[k],

[0094] and blocks 5b, 6, and 8 be repeated for a different value of k, using the updated preliminary estimate.

[0095] In particular, the estimation may be blocks 5b, 6, and 8 may be repeated in sequence for successive values of frequency indices k=1, 2, . . . K.sub.f−1. In that case, in the case of a determination of H.sub.est[k]=−H.sub.prelim[k], i.e., a phase reversal required, for a frequency index k=m, all subsequent preliminary estimates may be phase reversed as well, i.e.,

H′.sub.prelim[m]=−H′.sub.prelim[m] for m≥k,

i.e., a sign-flip of all measurements stating at frequency index m. If two such sign changes are observed at say m and n with m<n, the values between m k<n are multiplied by −1 and the remainder by (−1).sup.2=1. For three or more jumps, the procedure is simply extended. To limit the number of multiplications, first a mask can be created to keep track of which values should be sign-flipped and which not.

[0096] FIG. 3a shows a simulated example of H.sub.sqrt[k] from one antenna at 80 different frequency channels that are spaced 1 MHz apart. The number markers are the frequency indices k.

[0097] It can be observed that between frequency 25 and 26 a large jump is present. In fact, the frequency 26 (and all subsequent) is phase-inversed. In this case if frequency 26 were to be flipped back with respect to the origin, its distance to frequency 25 would be much smaller and thus more reasonable compared to the distance of other successive frequency pairs. Once the phase reversal is detected, the samples at frequency 26 and all subsequent frequencies may have their phase reversed.

[0098] FIG. 3b shows the final estimation H.sub.est[k] determined according to the above. Compared to H.sub.sqrt[k] of FIG. 3a, the large jump between frequency indices 25 and 26 has been removed, indicating an improved reconstruction of the one-way frequency domain channel response.

[0099] For the optional transformation of block 5a, an average phase advance between consecutive frequency indices k may be calculated as:

[00001]$\Delta {\varphi }_{\mathrm{ave}}\left[k\right]=\frac{{.Math.}_{k=1}^{K_f-1}\angle \left(X\left[k+1\right]X^{*}\left[k\right]\right)}{2\left(K_f-1\right)}$

[0100] where an asterisk denotes the complex conjugate and L denotes taking the argument of the complex number, i.e., the angle function.

[0101] Thus, the average phase advance is calculated by taking the argument of the product of the complex number representing the value proportional to the two-way frequency domain channel response for a frequency and the conjugate of the complex number representing the value proportional to the two-way frequency domain channel response for an adjacent frequency, summing over frequencies, and dividing by two times the number of frequency steps. In the case of non-uniform frequency spacing between successive frequency indices k, the formula may be modified accordingly.

[0102] Alternatively, as only the argument/phase of X[k] is used, the magnitude of this value may be omitted from the calculation, only using the argument of X[k].

[0103] Then, the transformation of block 5a may be defined as

H′.sub.prelim[k]=H.sub.prelim[k]e.sup.−j(k−1)Δϕ.sup.ave.sup.[k]

H′.sub.est[k]=H.sub.est[k]e.sup.−j(k−1)Δϕ.sup.ave.sup.[k]

[0104] This results in a down-mixing, removing an overlaid inherent phase advance between adjacent frequency indices. Thus, each such representation H.sub.sqrt[k] and H′.sub.est[k] of a respective quantity H.sub.sqrt[k] and H.sub.est[k] is a frequency-dependent transformation compensating for an inherent phase advance between adjacent frequencies.

[0105] FIG. 3c shows the transformed H′.sub.prelim[k]=H′.sub.sqrt[k] based on the H.sub.prelim[k]=H.sub.sqrt[k] of FIG. 3a. As can be seen, the inherent overlaid phase rotation between successive frequency indices is removed. This increases the likelihood of an accurate predicted estimate at block 5b. The predicted estimate at block 5b may either be an adjacent estimate according to the above, or an extrapolation as will be exemplified below.

[0106] As mentioned above, the predicted estimate of block 5b, for a frequency index k, may be an extrapolation from two or more estimates for two or more respective adjacent frequencies to said frequency. Extrapolation may be performed using estimates H′.sub.prelim[k] as transformed by the transformation described above, or directly on the preliminary estimates H.sub.prelim[k], i.e., with H′.sub.prelim[k]=H.sub.prelim[k].

[0107] To simplify notation, define a vector

h′=[H′.sub.prelim[0],H′.sub.prelim[1], . . . ,H′.sub.prelim[K.sub.ƒ]]

and let h′.sub.k denote the kth element of h′, and h′.sub.a:b, a sub-vector of h′ starting from the ath element and ending, and including the bth element, in that specific order.

[0108] Extrapolation may be performed at an order M, where M signifies the number of adjacent points used for the extrapolation, where M=2, 3, 4 . . . .

[0109] An extrapolation function ƒ(x), as known per se, may be defined, where x signifies an M-dimensional vector of complex values from which the extrapolations should be performed.

[0110] For example, ƒ(x) may be a linear extrapolation function, corresponding to order M=2. Such a function may be written

ƒ([a b])=b+(b−a)=2b−a

and will be used below.

[0111] As another example, ƒ(x) may be a cubic extrapolation function, corresponding to M=3.

[0112] The predicted estimate of block 5b (see above) may then be calculated as

H′.sub.pred[k]=H′.sub.pred,L[k]=ƒ(h′.sub.k−M:k−1),

[0113] and the first metric distance d.sub.1 and the second metric distance d.sub.2 calculated at block 6 as described above.

[0114] Here, the two or more respective adjacent frequencies to said frequency are lower than the frequency corresponding to frequency index k, as signified by the sub-script L.

[0115] Additionally, extrapolation may be double-sided, both from below and from above in frequency. Then, a second predicted estimate of the representation of the value proportional to the one-way frequency domain channel response at frequency index k−1 may be calculated as

H′.sub.pred,H[k][k−1]=ƒ(h′.sub.k+M−1:−1:k),

where “:−1:” indicates a reversal of the elements of the sub-vector. Thus, H.sub.pred,H [k−1] is based on an extrapolation from one or more estimates for frequency index k and one or more respective frequency indices k+1, k+2, . . . , corresponding to frequencies adjacent to and higher than the frequency corresponding to frequency index k, as signified by the sub-script H.

[0116] Similar to the first metric distance d.sub.L and the second metric distance d′.sub.L a third metric distance d.sub.H may be calculated, still at block 6, between the second predicted estimate H′.sub.pred,H[k−1] and the representation H′.sub.prelim[k−1] of the preliminary estimate as

d.sub.H=∥H′.sub.prelim[k−1]−H′.sub.pred,H[k−1]∥

[0117] and a fourth metric distance d′.sub.H may be calculated between the second predicted estimate H′.sub.pred,H[k−1] and a phase reversal—H′.sub.prelim[k−1] of the representation of the preliminary estimate as

d′.sub.H=∥H′.sub.prelim[k−1]−H′.sub.pred,H[k−1]∥=∥H′.sub.prelim[k−1]+H′.sub.prelim[k−1]+H′.sub.pred,H[k−1]∥,

where the condition d′.sub.H<d.sub.H indicates an incorrect sign in the preliminary estimate H.sub.prelim[k] as it indicates a sign flip between H.sub.prelim[k−1] and H.sub.prelim[k].

[0118] The information from comparing, respectively, d.sub.L and d′.sub.L, and d.sub.H and d′.sub.H may be combined. Thus, for example, at block 8, the final estimation may be

H.sub.est[k]=−H.sub.prelim[k]

if and only if d′.sub.L<d.sub.L and d′.sub.H<d.sub.H, and

H.sub.est[k]=H.sub.prelim[k]

otherwise.
Thus, in this case of double-sided extrapolation, the final estimate H.sub.est[k] of the value proportional to said one-way frequency domain channel response is based on comparison of the first metric distance d.sub.L, the second metric distance d′.sub.L the third metric distance d.sub.H, and the fourth metric distance d′.sub.H.

[0119] The functioning of the extrapolation from both below and above in frequency may be better understood with reference to FIGS. 3b, 3c, and 4, illustrating a limitation of reconstruction without extrapolation.

[0120] With reference to FIG. 3b, even though the jump at frequency index 26 is corrected successfully, frequency index 7 was mistakenly phase reversed, as is evident from the apparent discontinuity of the otherwise smooth progression of the direction of the tangent of the resulting curve, due to that points small magnitude compared to its distance to the point corresponding to the previous frequency index 6. Such a phenomenon may for example occur when experiencing deep fading at certain frequencies, due to multipath propagation. FIG. 3c shows that this discontinuity is present as well, in this example, after the transformation to compensate for the inherent phase advance.

[0121] FIG. 4 is a close-up look of the area around the point corresponding to frequency index 7 in FIG. 3c and illustrates linear extrapolation according to the above for frequency index k=7.

[0122] Shown with plus symbols and frequency indices are the transformed representations H′.sub.prelim[k] of the preliminary estimates H.sub.prelim[k].

[0123] A phase reversal—H′.sub.prelim[7] (corresponding to reflection in the origin) of the representation of the preliminary estimate H′.sub.prelim [7] for frequency index 7 is marked 7′. In the same way, a phase reversal—H′.sub.prelim[6] of the representation of the representation of the preliminary estimate H′.sub.prelim[6] for frequency index 6 is marked 6′.

[0124] Further, shown with an unfilled (white) circle is a prediction H′.sub.pred [7] based on a linear extrapolation from lower frequencies, viz., from preliminary estimates H′.sub.prelim [5] and H′.sub.prelim[6]

[0125] Metric distances in the complex plane d.sub.L—between H′.sub.prelim [7](“7”) and H′.sub.pred,L [7] (unfilled circle)—and d′.sub.L between—H′.sub.prelim [7](“7”) and H′.sub.pred,L [7] (unfilled circle) are further shown.

[0126] Since, in this example, d′.sub.L<d.sub.L, the phase-inversed preliminary estimate −H.sub.prelim [7] becomes the final estimate, i.e., H.sub.est[7]=−H.sub.prelim [7].

[0127] Further, extrapolation from higher frequency is shown. The prediction H′.sub.pred,H [6] from a linear extrapolation from preliminary estimates H′.sub.prelim[8] and H.sub.prelim [7] is shown with a filled (black circle).

[0128] Metric distances in the complex plane d.sub.H—between H′.sub.prelim [6] (“6”) and H′.sub.pred,H [6] (filled circle)—and d′.sub.H between −H.sub.prelim[6] (“6”) and H′.sub.pred,H [6] (filled circle) are further shown.

[0129] Thus, in this example, d′.sub.H<d.sub.H. This further indicated that a phase reversal has occurred for the preliminary estimate between frequency indices 6 and 7, indicating the need to for a phase reversal in the final estimate for frequency index 7 in compensation.

[0130] Thus, with both d′.sub.L<d.sub.L and d′.sub.H<d.sub.H, also when performing double-sided extrapolation, H.sub.est[7]=−H.sub.prelim [7].

[0131] Finally, in block 10 of FIG. 3, the distance between the first and the second radio signal transceiver may be determined based on the final estimates. E.g. an inverse fast Fourier transform (IFFT), as is known per se, can be used but also more advanced signal processing techniques typically referred to as super-resolution algorithms, as also are known per se, as described in, e.g., [0132] Schmidt: IEEE Transactions on Antennas and Propagation, Vol AP-34, No. 3, pp. 276-280, March 1986; [0133] Sakar: IEEE Antennas and Propagation Magazine, Vol. 37, No. 1, pp. 48-55, February 1995; [0134] Roy: IEEE Transactions on Acoustics, Speech, and Signal Processing, Vol. 37, No. 7, July 1989.

[0135] The reconstructed one-way frequency-domain channel response H[k] allows most ranging algorithms to mitigate more interference from multipath, as the order of the problem/number of components is reduced. In the presence of multipath, the number of components interfering with the estimation of the delay of the line-of-sight (LOS) component will be reduced and ranging and localization will be more accurate.

[0136] A computer program product comprising a computer-readable medium may store computer-readable instructions such that when executed on a processing unit the computer program product will cause the processing unit to perform the method according to the above.

[0137] The method may be performed in a processing unit, which may be arranged in a device A, B or C as discussed above.

[0138] The processing unit may be implemented in hardware, or as any combination of software and hardware. At least part of the functionality of the processing unit may, for instance, be implemented as software being executed on a general-purpose computer. The system may thus comprise one or more processing units, such as a central processing unit (CPU), which may execute the instructions of one or more computer programs in order to implement desired functionality.

[0139] The processing unit may alternatively be implemented as firmware arranged e.g. in an embedded system, or as a specifically designed processing unit, such as an Application-Specific Integrated Circuit (ASIC) or a Field-Programmable Gate Array (FPGA).

[0140] The correlation properties for the one-way frequency domain channel response for adjacent frequencies will naturally vary depending on the exact environment. There may be a range of frequency step sizes where the method disclosed herein works increasingly well as the frequency stepping is reduced, but where no hard upper limit of applicability can be defined.

[0141] The concept of coherence bandwidth is a statistical measurement that is approximately the maximum frequency interval over which two signals at two frequencies experience correlated amplitude fading. An approximation of the coherence bandwidth over which the amplitude correlation is lower than 0.5 is

[00002]$B_c=\frac{1}{5{\sigma }_{T_m}}$

[0142] where σ.sub.T.sub.m is the root mean square (RMS) delay spread (Goldsmith, Andrea. Wireless communications. Cambridge university press, 2005.)

[0143] According to a measurement campaign at 2.4 GHz in a 7.8 m-by-10 m room where benches and laboratory equipment scatter around, the RMS delay spread ranges from 20 ns to 30 ns depending on the relative distance between Tx and Rx (Zepernick, H. J., & Wysocki, T. A. Multipath channel parameters for the indoor radio at 2.4 GHz ISM band. In 1999 IEEE 49th Vehicular Technology Conference, May 1999, Vol. 1, pp. 190-193). This indicates a coherence bandwidth in such an environment be around 8 MHz (with σ.sub.T.sub.m=25 ns).

[0144] A typical ranging and direction-finding system may operate in the ISM band with a frequency step in the order of 1 MHz. According to the above, this this frequency step is small enough that the one-way frequency domain channel responses are correlated.

[0145] Additionally, the phase of frequency responses that are at different frequencies are correlated as well, because the slope of the phase along frequency is proportional to the distance between the initiator and reflector in a line-of-sight (LOS) channel. The amplitude coherency and phase coherency across frequency samples of the wireless channel when observed at every 1 MHz ensures that the complex frequency response rotates progressively rather than arbitrarily.

[0146] In the following, results validating the channel reconstruction of the one-way frequency domain channel response according to the present disclosure, including the transformation and double-sided linear extrapolation as detailed above.

[0147] Multi-path channels were generated by means of ray tracing. In the ray tracer, a 11 m-by-7 m meeting room without furniture was defined, as illustrated in FIG. 5. An initiator was put in one of the corners while the reflector was moved around the room at 112 distinct locations forming the shape of an Arabic numeral 8. The scene is shown in FIG. 5. The simulation contained up to 3rd order reflections. Random phase offsets were selected uniformly randomly from 0 to 27π radians and added to the channel response each time the carrier frequency was switched. The phase offset was kept constant when switching the direction of transmission. White Gaussian noise was added such that the SNR per antenna was 20 dB.

[0148] An indicator of reconstruction reliability was defined in the form of a reconstructed-channel quality indicator Q, describing the similarity between the reconstructed channel and the actual channel. It is defined by:

[00003]$Q=.Math.\frac{h_{est}^{H}h}{\sqrt{h^{H}h}\sqrt{h_{est}^H{h}_{est}}}.Math.$

[0149] where h.sub.est and h are column vectors, which are the estimated channel and the actual channel vectors, respectively, and .sup.H denotes the operation of conjugate transpose. The closer Q is to 1, the better the reconstruction matches the actual channel.

[0150] FIGS. 6a, 6b, 6c, and 6d respectively show the quality indicator Q for Rician-K values 10, 2, 5, and 1. The higher the Rician-K value, the stronger the line-of-sight component is with respect to multipath propagation. Each of the figures shows the Q value for a number of simulated radio channels as described above.

[0151] One can see that for high K values, the channel reconstruction is flawless. If the Rician K value reduces, performance of the algorithm somewhat reduces, while still a good reconstruction is realized for a significant part of the channels.

[0152] In the above the inventive concept has mainly been described with reference to a limited number of examples. However, as is readily appreciated by a person skilled in the art, other examples than the ones disclosed above are equally possible within the scope of the inventive concept, as defined by the appended claims.

[0153] For example, the extrapolation can be done linearly, as exemplified in this disclosure, or in more sophisticated manners, e.g. using higher-order, non-linear extrapolation.

[0154] Instead of calculating an average phase advance transformation compensating for the inherent phase advance between adjacent frequencies may be provided through a tracking algorithm.