Time difference of arrival (TDOA) estimation method, apparatus, electronic device, and readable storage medium

Abstract

The application provides a time difference of arrival (TDOA) estimation method, device, electronic equipment and readable storage medium. The method comprises: obtaining an amplitude spectrum and an unwrapped phase difference spectrum of an i-th speech frame, i being a positive integer; obtaining frequency points on the left side of mirror frequency points in the amplitude spectrum to obtain a first frequency point set of the i-th speech frame; determining a first phase difference slope spectrum corresponding to the i-th speech frame according to a first amplitude set corresponding to the first frequency point set in the amplitude spectrum and a first phase difference set corresponding to the first frequency point set in the unwrapped phase difference spectrum; and determining a TDOA of the i-th speech frame according to a first smoothed phase difference frequency spectrum corresponding to the first phase difference frequency spectrum.

Description

Technical Field

[0001] The embodiments of the present application relate to the field of communication technology, and in particular to a method, device, electronic device, and readable storage medium for estimating time difference of arrival (TDOA). Background Art

[0002] Time Difference of Arrival (TDOA) estimation is one of the basic tasks in speech signal processing, and TDOA can be used to locate the speaker.

[0003] Currently, the most commonly used TDOA estimation method is based on cross-correlation. This method uses a sampling point-by-sample search and comparison method, selecting the sampling point interval corresponding to the maximum cross-correlation function as the TDOA estimate. During the implementation of this application, the inventors discovered that the existing technology has at least the following problems: the TDOA estimate can only be an integer multiple of the sampling period, which indicates that the existing TDOA estimation accuracy is low. Summary of the Invention

[0004] The embodiments of the present application provide a method, device, electronic device and readable storage medium for estimating the time difference of arrival (TDOA), which can solve the problem in the prior art that the estimation resolution of TDOA is limited by the sampling period, resulting in low TDOA estimation accuracy.

[0005] To solve the above problems, this application is implemented as follows:

[0006] In the first aspect, an embodiment of the present application provides a TDOA estimation method, including: obtaining the amplitude spectrum and unwarped phase difference spectrum of the i-th voice frame, where i is a positive integer; obtaining the frequency point located to the left of the mirror frequency point in the amplitude spectrum to obtain a first frequency point set of the i-th voice frame; determining the first phase difference slope spectrum corresponding to the i-th voice frame based on the unwarped phase difference set corresponding to the first frequency point set in the unwarped phase difference spectrum, and the first amplitude set corresponding to the first frequency point set in the amplitude spectrum; determining the TDOA of the i-th voice frame based on the first smoothed phase difference frequency spectrum corresponding to the first phase difference frequency spectrum.

[0007] In the second aspect, an embodiment of the present application also provides a TDOA estimation device, including: a first acquisition module, used to obtain the amplitude spectrum and unwarped phase difference spectrum of the i-th voice frame, where i is a positive integer; a second acquisition module, used to obtain the frequency points located to the left of the mirror frequency points in the amplitude spectrum to obtain the first frequency point set of the i-th voice frame; a first determination module, used to determine the first phase difference slope spectrum corresponding to the i-th voice frame based on the unwarped phase difference set corresponding to the first frequency point set in the unwarped phase difference spectrum, and the first amplitude set corresponding to the first frequency point set in the amplitude spectrum; a second determination module, used to determine the TDOA of the i-th voice frame based on the first smoothed phase difference frequency spectrum corresponding to the first phase difference frequency spectrum.

[0008] In a third aspect, an embodiment of the present application further provides an electronic device comprising a processor, a memory, and a program or instruction stored in the memory and executable on the processor, wherein the program or instruction, when executed by the processor, implements the TDOA estimation method as described in the first aspect.

[0009] In a fourth aspect, an embodiment of the present application further provides a readable storage medium having a program or instruction stored thereon, which, when executed by a processor, implements the TDOA estimation method as described in the first aspect.

[0010] In a fifth aspect, an embodiment of the present application provides a chip, which includes a processor and a communication interface, the communication interface and the processor are coupled, and the processor is used to run programs or instructions to implement the method described in the first aspect.

[0011] In an embodiment of the present application, when estimating the TDOA of each speech frame, the amplitude spectrum and unwrapped phase difference spectrum of the speech frame are first obtained, and the first frequency point set of the speech frame is screened out based on the mirror frequency point. Afterwards, the first phase difference slope spectrum corresponding to the speech frame is determined using the unwrapped phase difference set corresponding to the first frequency point set in the unwrapped phase difference spectrum, and the first amplitude set corresponding to the first frequency point set in the amplitude spectrum. The TDOA of the speech frame is determined based on the first smoothed phase difference frequency spectrum corresponding to the first phase difference frequency spectrum. It can be seen that the estimation of TDOA in the embodiment of the present application eliminates the limitation of the sampling period on the estimation resolution, thereby improving the estimation accuracy of TDOA. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Figure 1 This is one of the flow charts of the TDOA estimation method provided in the embodiment of the present application;

[0013] Figure 2 This is one of the effect diagrams provided in the embodiment of the present application;

[0014] Figure 3 This is the second effect diagram provided by the embodiment of the present application;

[0015] Figure 4 This is the third effect diagram provided by the embodiment of the present application;

[0016] Figure 5 This is the fourth effect diagram provided by the embodiment of the present application;

[0017] Figure 6 This is the fifth effect diagram provided by the embodiment of the present application;

[0018] Figure 7 This is the sixth effect diagram provided by the embodiment of the present application;

[0019] Figure 8 This is the second flowchart of the TDOA estimation method provided in the embodiment of the present application;

[0020] Figure 9 This is the seventh effect diagram provided by the embodiment of the present application;

[0021] Figure 10 This is the eighth effect diagram provided by the embodiment of the present application;

[0022] Figure 11 is a structural diagram of a TDOA estimation device provided in an embodiment of the present application;

[0023] Figure 12 This is one of the structural diagrams of the electronic device provided in the embodiment of the present application;

[0024] Figure 13 This is the second structural diagram of the electronic device provided in the embodiment of the present application. DETAILED DESCRIPTION

[0025] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are part of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0026] The terms "first," "second," and the like in the specification and claims of this application are used to distinguish similar objects, and are not used to describe a specific order or precedence. It should be understood that the terms used in this manner are interchangeable where appropriate, so that the embodiments of this application can be implemented in an order other than that illustrated or described herein, and that the objects distinguished by "first," "second," and the like are generally of the same type, and do not limit the number of objects; for example, the first object can be one or more. In addition, the term "and / or" in the specification and claims refers to at least one of the connected objects, and the character " / " generally indicates that the objects connected are in an "or" relationship.

[0027] The TDOA estimation method provided in the embodiments of the present application is described in detail below with reference to some embodiments and their application scenarios in conjunction with the accompanying drawings.

[0028] The TDOA estimation method of the embodiment of the present application can be performed by an electronic device. The electronic device includes at least two microphones (mic). For ease of understanding, the following example is described as an electronic device including a first microphone and a second microphone, but the number of microphones included in the electronic device is not limited thereto.

[0029] In practical applications, electronic devices may be mobile phones, tablet personal computers, laptop computers or notebook computers, personal digital assistants (PDAs), handheld computers, netbooks, ultra-mobile personal computers (UMPCs), mobile Internet devices (MIDs), augmented reality (AR) / virtual reality (VR) devices, robots, wearable devices (Wearable Devices), vehicle-mounted equipment (VUEs), pedestrian terminals (PUEs), smart homes (home appliances with wireless communication functions, such as refrigerators, televisions, washing machines or furniture, etc.), and other terminal-side devices. Wearable devices include: smart watches, smart bracelets, smart headphones, smart glasses, smart jewelry (smart bracelets, smart bracelets, smart rings, smart necklaces, smart anklets, smart anklets, etc.), smart wristbands, smart clothing, game consoles, etc.

[0030] In the following embodiments, the i-th speech frame may be understood as any speech frame corresponding to the collected speech signal. For each speech frame, the TDOA estimation method of the embodiment of the present application may be used to estimate its TDOA.

[0031] See also Figure 1 , Figure 1 This is one of the flow charts of the TDOA estimation method provided in the embodiment of the present application. Figure 1 As shown, the TDOA estimation method of the embodiment of the present application may include the following steps:

[0032] Step 101: Obtain the amplitude spectrum and unwarped phase difference spectrum of the i-th speech frame, where i is a positive integer.

[0033] In a specific implementation, after the speech signal is framed to obtain the i-th speech frame, a Fourier transform can be performed on the i-th speech frame to obtain the amplitude spectrum (ampCur) and phase difference spectrum (phaseDiff) of the i-th speech frame. The ampCur includes each frequency point corresponding to the i-th speech frame and the amplitude value corresponding to each frequency point. The phaseDiff includes each frequency point corresponding to the i-th speech frame and the phase difference value of two microphones (or two channels) corresponding to each frequency point.

[0034] Afterwards, an unwrapping operation is performed on the phase difference spectrum to obtain an unwrapped phase difference spectrum (diffUnwrap). diffUnwrap has good flatness and consistency characteristics and can eliminate the influence of wrapping. In this way, using diffUnwrap to determine the TDOA of the i-th speech frame can improve the estimation accuracy of TDOA.

[0035] In an optional implementation, the unwrapping operation may include: taking the remainder of 2π(pi) on phaseDiff, and then performing a sine transform and an arcsine transform in sequence, but is not limited thereto.

[0036] The effects of ampCur, phaseDiff and diffUnwrap can be seen in Figure 2 .in, Figure 2 (a) shows the amplitude spectrum ampCur of a channel (channels correspond to microphones one-to-one) in three adjacent frames; Figure 2 (b) shows the phaseDiff of three adjacent frames; Figure 2 (c) shows the diffUnwrap of three adjacent frames. Figure 2 It can be seen that diffUnwrap shows good flatness characteristics in the sub-band area with strong amplitude.

[0037] Step 102: Obtain the frequency points on the left side of the mirror image frequency point in the amplitude spectrum to obtain a first frequency point set of the i-th speech frame.

[0038] In a specific implementation, the electronic device can use the mirror frequency point (floorBin4mirror) to filter out the frequency points in ampCur that are located to the left of floorBin4mirror, that is, the frequency points that are smaller than floorBin4mirror, to obtain a first frequency point set (iBinBig). In this case, iBinBig includes the frequency points in the amplitude spectrum that are located to the left of the mirror frequency point.

[0039] The mirror frequency point (floorBin4mirror) can be calculated based on the dual-microphone distance (distMic), the sampling frequency (fs), the Fourier transform length (lenFft), etc. In some embodiments, the i-th voice frame is collected by the first microphone and the second microphone, and the amplitude spectrum is obtained by performing Fourier transform on the i-th voice frame; before obtaining the frequency point located to the left of the mirror frequency point in the amplitude spectrum to obtain the first frequency point set of the i-th voice frame, the method further includes: determining the mirror frequency point based on the distance between the first microphone and the second microphone, the sampling frequency of the first microphone, and the Fourier transform length of the Fourier transform. In specific implementation, floorBin4mirror can be calculated by the following formula: floorBin4mirror = velocity × lenFft / (fs × distMic), but is not limited to this.

[0040] Through the above method, when determining the first frequency point set, only the frequency points to the left of the loorBin4mirror frequency point are considered. This can avoid serious errors caused by diffUnwrap mirror wrapping in scenarios where TDOA is too large, thereby improving the estimation accuracy of TDOA.

[0041] Step 103: Determine a first phase difference slope spectrum corresponding to the i-th speech frame based on the unwarped phase difference set corresponding to the first frequency point set in the unwarped phase difference spectrum and the first amplitude set corresponding to the first frequency point set in the amplitude spectrum.

[0042] After determining iBinBig, you can first determine the unwrapped phase difference set corresponding to iBinBig in diffUnwrap (diffUnwrap(iBinBig) or diffBig); and the first amplitude set corresponding to iBinBig in ampCur (ampCur(iBinBig)). diffBig has obvious equal efficiency distribution characteristics, see for details. Figure 3 .

[0043] Figure 3 (a) shows ampCur of three adjacent frames; Figure 3(b) shows the diffUnwrap of three adjacent frames; Figure 3 (c) shows the diffBig of three adjacent frames. Figure 3 It can be seen that diffBig shows a good equal slope distribution feature in the sub-band area with stronger amplitude.

[0044] Afterwards, the first phase difference slope spectrum (slopeBigCur) is determined based on diffUnwrap(iBinBig) and ampCur(iBinBig). In an optional implementation, diffUnwrap(iBinBig) can be weighted averaged by ampCur(iBinBig) to obtain slopeBigCur. In specific implementation, for each frequency point in iBinBig, its slope value can be calculated by the following formula: slope value of frequency point = (amplitude value corresponding to frequency point × unwrapped phase difference corresponding to frequency point) / frequency point. Since diffBig exhibits good equal slope distribution characteristics in the sub-band area with strong amplitude, estimating TDOA based on diffBig can further improve the estimation accuracy of TDOA.

[0045] Step 104: Determine the TDOA of the i-th speech frame according to the first smoothed phase difference frequency spectrum corresponding to the first phase difference frequency spectrum.

[0046] During specific implementation, after slopeBigCur is obtained, historical smoothing may be performed on slopeBigCur to obtain a first smoothed phase difference frequency spectrum (slopeBigSmth).

[0047] The effect diagram of slopeBigCur and slopeBigSmth can be seen in Figure 4 . Figure 4 (a) shows the spectrogram of a certain channel; Figure 4 (b) shows slopeBigCur and slopeBigSmth. Figure 4 It can be seen that slopeBigSmth can approach a stable value well, such as Figure 4 medium, tending towards 0.01.

[0048] In one implementation, the stable value approached by slopeBigSmth can be directly mapped to TDOA. In this case, the TDOA of the i-th speech frame can be determined by looking up the mapping relationship between the stable value and TDOA.

[0049] In other implementations, the electronic device may determine the TDOA of the i-th voice frame in combination with slopeBigSmth and other information. For details, please refer to the following related description, which will not be repeated here.

[0050] The TDOA estimation method of this embodiment, when estimating the TDOA of each speech frame, first obtains the amplitude spectrum and the unwrapped phase difference spectrum of the speech frame, and screens out the first frequency point set of the speech frame according to the amplitude threshold value corresponding to the amplitude spectrum and the mirror frequency threshold value. Afterwards, the first phase difference slope spectrum corresponding to the speech frame is determined by using the unwrapped phase difference set corresponding to the first frequency point set in the unwrapped phase difference spectrum and the first amplitude set corresponding to the first frequency point set in the amplitude spectrum. The TDOA of the speech frame is determined according to the first smoothed phase difference frequency spectrum corresponding to the first phase difference frequency spectrum. It can be seen that the TDOA estimation of the embodiment of the present application eliminates the limitation of the sampling period on the estimation resolution, thereby improving the estimation accuracy of TDOA.

[0051] In addition, it can be seen from the above content that the estimation of TDOA in the embodiment of the present application can be achieved through four arithmetic operations and historical smoothing processing, thereby reducing the estimation complexity of TDOA.

[0052] In some embodiments, the frequency points included in iBinBig can be directly determined based on floorBin4mirror. In other embodiments, the frequency points included in iBinBig can be further determined in combination with the amplitude threshold value (floorBigAmp), as described below:

[0053] Acquiring the frequency points on the left side of the mirror frequency point in the amplitude spectrum to obtain the first frequency point set of the i-th speech frame may include:

[0054] Determining a threshold value corresponding to the i-th speech frame according to the amplitude mean and the amplitude maximum of the amplitude spectrum;

[0055] Performing historical smoothing on the threshold value to obtain an amplitude threshold value corresponding to the amplitude spectrum;

[0056] Obtaining frequency points that meet a first condition in the amplitude spectrum to obtain a first frequency point set of the i-th speech frame;

[0057] The first condition includes: being located on the left side of the mirror frequency point, and the corresponding amplitude value being greater than the amplitude threshold value.

[0058] The amplitude threshold value (floorBigAmp) corresponding to ampCur can be calculated based on the mean value (meanAmp) and the maximum value (peakAmp) of ampCur.

[0059] In a specific implementation, the threshold value (floorBigCur) may be calculated using the following formula: floorBigCur=meanAmp+0.3×(peakAmp−meanAmp), but is not limited thereto.

[0060] In one embodiment, the electronic device can first use floorBigAmp to filter out a frequency point set in ampCur whose amplitude value is greater than floorBigAmp, and then use floorBin4mirror to filter out the frequency points to the left of floorBin4mirror in the frequency point set, that is, the frequency points less than floorBin4mirror, to obtain the first frequency point set (iBinBig).

[0061] In another embodiment, the electronic device can first use floorBin4mirror to filter out the frequency points in ampCur located to the left of floorBin4mirror, that is, the frequency points less than floorBin4mirror, to obtain a frequency point set, and then use floorBigAmp to filter out the frequency points in the frequency point set whose corresponding amplitude values ​​are greater than floorBigAmp to obtain iBinBig.

[0062] Through the above method, when determining the first frequency point set, only the frequency points located to the left of the loorBin4mirror frequency point and with amplitude values ​​greater than floorBigAmp are considered. In this way, on the one hand, serious errors caused by diffUnwrap mirror wrapping in the scenario where TDOA is too large can be avoided, and on the other hand, information with weak anti-interference can be filtered out, thereby improving the estimation accuracy of TDOA.

[0063] In some embodiments, determining the TDOA of the i-th voice frame according to the first smoothed phase difference slope spectrum corresponding to the first phase difference slope spectrum may include:

[0064] determining, based on a first smoothed phase difference slope spectrum corresponding to the first phase difference slope spectrum, frequency points at which mirror warping occurs in the dewarped phase spectrum, to obtain a second frequency point set of the i-th speech frame;

[0065] de-imaging the dewarped phase difference spectrum according to the second frequency point set to obtain a de-imaging phase difference spectrum;

[0066] determining, according to the amplitude mean corresponding to the amplitude spectrum, a third frequency point set of the i-th speech frame, wherein the amplitude value corresponding to each frequency point in the third frequency point set is greater than the amplitude mean;

[0067] Determining a second phase difference slope spectrum corresponding to the i-th speech frame according to a de-imaging phase difference set corresponding to the third frequency point set in the de-imaging phase difference spectrum and a second amplitude set corresponding to the first frequency point set in the amplitude spectrum;

[0068] The TDOA of the i-th voice frame is determined according to the first smoothed phase difference slope spectrum and a second smoothed phase difference slope spectrum corresponding to the second phase difference slope spectrum.

[0069] In this embodiment, the electronic device may first calculate the frequency points where mirror wrapping occurs in diffUnwrap based on slopeBigSmth to obtain a second frequency point set (iBinMirror) of the i-th voice frame.

[0070] In an optional implementation, the amplitude spectrum is obtained by performing Fourier transform on the i-th speech frame;

[0071] Determining, based on the first smoothed phase difference slope spectrum corresponding to the first phase difference slope spectrum, the frequency points where mirror warping occurs in the dewarped phase spectrum to obtain the second frequency point set of the i-th speech frame may include:

[0072] determining a first frequency point according to a maximum slope value in a first smoothed phase difference slope spectrum corresponding to the first phase difference slope spectrum and a Fourier transform length of the Fourier transform;

[0073] The frequency point where the mirror wrapping occurs is determined according to the first frequency point and the maximum slope value, to obtain a second frequency point set of the i-th voice frame.

[0074] In a specific implementation, the electronic device can calculate the amplitude spectrum length (lenAmp) according to the Fourier transform length (lenFft). In an optional implementation, lenAmp=lenFft / 2+1.

[0075] After the electronic device determines the maximum slope value (slopeBig) in slopeBigSmth and lenAmp, in one implementation, the first frequency point (nBinMirror) can be calculated by the following calculation formula: nBinMirror = floor(lenAmp*slopeBig / pi), where floor represents a rounding-down function.

[0076] In one implementation, the frequency point at which mirror wrapping occurs may be calculated using the following formula: iBinMirror=round(pi / slopeBig×(1:nBinMirror)′), where round represents a rounding function.

[0077] Then, iBinMirror is used to de-mirror the diffUnwrap to obtain a de-mirrored phase difference spectrum (diffMirror). In an optional implementation, the de-mirroring operation may include: performing a 2pi multiple flip on the diffUnwrap based on iBinMirror, but is not limited thereto.

[0078] The effect of diffMirror can be seen in Figure 5 . Figure 5 (a) shows ampCur of three adjacent frames; Figure 5 (b) shows the diffUnwrap of three adjacent frames; Figure 5 (c) shows the diffMirror of three adjacent frames. Figure 5 It can be seen that there is no obvious mirror wrapping error in diffMirror. Therefore, calculating TDOA based on diffMirror can improve the estimation accuracy of TDOA.

[0079] The amplitude mean (ceilSmallAmp) corresponding to ampCur may be determined based on meanAmp of ampCur. In an optional implementation, meanAmp may be historically smoothed to obtain ceilSmallAmp.

[0080] After ceilSmallAmp is determined, a frequency point set in ampCur having an amplitude greater than ceilSmallAmp may be determined as the third frequency point set (iBinNoSmall).

[0081] Then, the de-mirroring phase difference set (diffMirror(iBinNoSmall)) corresponding to iBinNoSmall in diffMirror and the second amplitude set (ampCur(iBinNoSmall)) corresponding to iBinNoSmall in ampCur are determined, and the second phase difference slope spectrum corresponding to the i-th speech frame is determined.

[0082] Then, the second phase difference slope spectrum (slopeNoSmallCur) is determined based on diffMirror(iBinNoSmall) and ampCur(iBinNoSmall). In an optional implementation, a weighted average of ampCur(iBinNoSmall) can be performed on diffMirror(iBinNoSmall) to obtain slopeNoSmallCur. In a specific implementation, for each frequency point in iBinNoSmall, its slope value can be calculated using the following formula: Slope value of frequency point = (amplitude value corresponding to frequency point × de-mirroring phase difference corresponding to frequency point) / frequency point.

[0083] SlopeNoSmallCur is historically smoothed to obtain the second smoothed phase difference slope spectrum (slopeNoSmallSmth). Compared with slopeBigSmth, slopeNoSmallSmth can better reflect the TDOA in the medium and high frequency range. The effect diagram of slopeNoSmallSmth can be seen in Figure 6 .

[0084] Figure 6 (a) shows ampCur; Figure 6 (b) shows slopeBigCur and slopeBigSmth; Figure 6 (c) shows slopeNoSmallCur and slopeNoSmallSmth. Figure 6 It can be seen that slopeNoSmallSmth can also approach a stable value well, which can directly map TDOA.

[0085] Therefore, in this embodiment, the electronic device can determine TDOA based on slopeBigSmth and slopeNoSmallSmth. In this way, compared with directly determining TDOA based on slopeBigSmth, TDOA in the high-frequency range is further considered, thereby further improving the estimation accuracy of TDOA.

[0086] In an optional implementation, the electronic device may determine an average value of a first TDOA mapped to a stable value approached by slopeBigSmth and a second TDOA mapped to a stable value approached by slopeNoSmallSmth as the TDOA of the i-th voice frame.

[0087] In another optional implementation, determining the TDOA corresponding to the i-th voice frame based on the first smoothed phase difference slope spectrum and the second smoothed phase difference slope spectrum corresponding to the second phase difference slope spectrum may include:

[0088] determining an average phase difference slope spectrum according to the first smoothed phase difference slope spectrum and a second smoothed phase difference slope spectrum corresponding to the second phase difference slope spectrum;

[0089] Determining an angle spectrum corresponding to the i-th speech frame according to the average phase difference slope spectrum;

[0090] The TDOA of the i-th speech frame is determined according to the angle spectrum.

[0091] During specific implementation, a weighted average of slopeBigSmth and slopeNoSmallSmth may be performed to obtain an average phase difference slope spectrum (slopeAvg).

[0092] Then, based on slopeAvg, the angle spectrum (angleCur) corresponding to the i-th speech frame is calculated. In an optional implementation, angleCur can be calculated based on information such as slopeAvg, distMic, velocity, and sampling frequency (fs). Specifically, angleCur can be calculated using the following formula: angleCur = 180 / pi × asin(slopeAvg × velocity × lenFft / (2 × pi × fs × distMic)), where asin represents the inverse sine function.

[0093] angleCur has the advantages of high precision and good noise resistance, etc. For details, please refer to Figure 7 .

[0094] Figure 7 (a) shows ampCur; Figure 7 (b) shows slopeBigCur, slopeNoSmallCur and slopeAvg; Figure 7 (c) shows angleCur. Figure 7 It can be seen that angleCur approaches the true incident angle (-20 degrees) with high accuracy and stability, indicating that the TDOA estimation is reliable.

[0095] The corresponding angleCur is then calculated based on information such as distMic and sound speed. The latter has the advantages of high precision and good noise resistance.

[0096] Based on this, the stable value of the angle spectrum can be directly determined as the TDOA of the i-th speech frame. In this way, the accuracy and noise resistance of the TDOA estimation can be improved.

[0097] In the embodiment of the present application, Figures 2 to 7 In the example, the incident angle (Angle) is -20 degrees, the Fourier transform length (lenFft) is 2048, the microphone spacing (distMic) is 0.2 meters (m), the sampling frequency (fs) is 16 kHz (kHz), the frame length lenFrame is 512, and the signal-noise ratio (SNR) is 100000 dB.

[0098] phaseDiff can also be called phase difference; diffUnwrap can also be called phase difference after unwrapping; ampCur can also be called current frame amplitude spectrum; floorBigCur can also be called strong amplitude lower limit; diffBig can also be called phase difference at strong amplitude frequency point; floorBin4mirror can also be called the lower limit of the frequency point where phase difference occurs in mirror image; slopeBigCur can also be called the slope of phase difference at strong amplitude frequency point of current frame; slopeBigSmth can also be called the historical smoothed value of the slope of phase difference at strong amplitude frequency point; iBinMirror can also be called the frequency point position where mirror wrapping occurs; dif fMirror can also be called the phase difference after de-mirroring; slopeNoSmallCur can also be called the slope of the phase difference of the non-weak amplitude frequency point of the current frame; slopeNoSmallSmth can also be called the historical smoothed value of the slope of the phase difference of the non-weak frequency point; slopeAvg can also be called the weighted average phase difference slope; angleCur can also be called the sound source incident angle of the current frame; meanAmp can also be called the mean of the amplitude spectrum of the current frame; peakAmp can also be called the maximum value of the amplitude spectrum of the current frame; iBinBig can also be called the strong amplitude frequency point set; iBinNoSmall can also be called the non-weak amplitude frequency point set.

[0099] It should be noted that the various optional implementation methods introduced in the embodiments of the present application can be implemented in combination with each other or separately if they do not conflict with each other, and the embodiments of the present application do not limit this.

[0100] For easier understanding, the following examples are provided:

[0101] The TDOA estimation method provided in this example has lower computational complexity than all existing methods, unlimited estimation resolution, and good noise immunity.

[0102] The TDOA estimation methods provided in this example include:

[0103] 1) Calculate the phase difference phaseDiff between channels, then take the modulo 2pi and the sine and arc sine changes. The resulting diffUnwrap has good flatness and consistency characteristics. The effect diagram can be seen in Figure 2 .in, Figure 2 (a) shows the amplitude spectrum ampCur of a channel (channels correspond to microphones one-to-one) in three adjacent frames; Figure 2 (b) shows the phaseDiff of three adjacent frames; Figure 2 (c) shows the diffUnwrap of three adjacent frames. Figure 2It can be seen that diffUnwrap shows good flatness characteristics in the sub-band area with strong amplitude.

[0104] Based on the analysis of this example, we find that diffUnwrap exhibits good flatness characteristics in sub-band regions with strong amplitudes.

[0105] 2) According to the mean and maximum values ​​of the amplitude spectrum ampCur of the current frame, a strong energy threshold floorBigCur is determined, and diffBig is extracted from diffUnwrap. The latter has obvious equal slope distribution characteristics. The effect example can be seen in Figure 3 .

[0106] Figure 3 (a) shows ampCur of the adjacent fifteen frames; Figure 3 (b) shows diffUnwrap of the adjacent fifteen frames; Figure 3 (c) shows diffBig of the adjacent fifteen frames.

[0107] Based on the analysis of this example, it is found that diffBig shows a good equal slope distribution feature in the sub-band area with strong amplitude.

[0108] 3) When determining diffBig, only the frequencies to the left of the floorBin4 mirror frequency are considered. This can avoid serious errors caused by diffUnwrap mirror wrapping in scenarios where TDOA is too large.

[0109] floorBin4mirror can be calculated based on the dual-microphone spacing distMic, the sampling frequency fs, and the Fourier transform length lenFft.

[0110] 4) Perform weighted ampCur averaging on diffBig to obtain slopeBigCur, and then perform historical smoothing to obtain slopeBigSmth, which has a one-to-one mapping relationship with TDOA. Figure 4 . Figure 4 (a) shows the spectrogram of a certain channel; Figure 4 (b) shows slopeBigCur and slopeBigSmth.

[0111] Based on the analysis of this example, it is found that slopeBigSmth can well approach a stable value, which can directly map TDOA.

[0112] 5) Determine the frequency point iBinMirror where mirror wrapping occurs in diffUnwrap based on slopeBigSmth, and then de-mirror diffUnwrap to obtain diffMirror. Figure 5 .

[0113] Figure 5 (a) shows ampCur of three adjacent frames; Figure 5 (b) shows the diffUnwrap of three adjacent frames; Figure 5 (c) shows the diffMirror of three adjacent frames.

[0114] Based on the analysis of this example, it is found that there is no obvious mirror warping error in diffMirror.

[0115] 6) Perform weighted ampCur averaging on diffMirror to obtain slopeNoSmallCur, and then perform historical smoothing to obtain slopeNoSmallSmth, which can better reflect the TDOA in the mid-high frequency range. Figure 6 .

[0116] Figure 6 (a) shows ampCur; Figure 6 (b) shows slopeBigCur and slopeBigSmth; Figure 6 (c) shows slopeNoSmallCur and slopeNoSmallSmth.

[0117] Based on the analysis of this example, it is found that slopeNoSmallSmth can also approach a stable value well, which can directly map TDOA.

[0118] 7) SlopeAvg is obtained by weighted average of slopeBigSmth and slopeNoSmallSmth, and then the corresponding angleCur is calculated based on information such as distMic and sound speed. The latter has the advantages of high precision and good noise resistance. Figure 7 .

[0119] Figure 7 (a) shows ampCur; Figure 7 (b) shows slopeBigCur, slopeNoSmallCur and slopeAvg; Figure 7 (c) shows angleCur.

[0120] Based on the analysis of this example, we find that angleCur approaches the true incident angle (-20 degrees) with high accuracy and stability, indicating that the TDOA estimation is reliable.

[0121] like Figure 8 As shown, the TDOA estimation method may include the following steps:

[0122] Step 801: The signal is divided into frames and then subjected to Fourier transform to obtain ampCur and phaseDiff.

[0123] Step 802: perform 2pi on phaseDiff and take the remainder, then perform sine transform and arcsine transform in sequence to obtain diffUnwrap.

[0124] Step 803: Calculate the meanAmp and the maximum peakAmp based on ampCur to obtain floorBigCur.

[0125] floorBigCur can be calculated using the following formula: floorBigCur=meanAmp+0.3*(peakAmp-meanAmp).

[0126] Step 804: Perform historical smoothing on meanAmp and floorBigCur to obtain ceilSmallAmp and floorBigAmp respectively.

[0127] Step 805: Select the frequency point set in ampCur that is greater than floorBigAmp to obtain iBinBig.

[0128] Step 806: Perform weighted averaging of ampCur(iBinBig) on ​​diffUnwrap(iBinBig) to obtain slopeBigCur.

[0129] Step 807: Perform historical smoothing on slopeBigCur to obtain slopeBigSmth.

[0130] Step 808: Calculate the frequency point position set iBinMirror where mirror wrapping occurs in diffUnwrap according to slopeBigSmth.

[0131] Step 809: Perform a 2pi multiple flip on diffUnwrap based on iBinMirror to obtain diffMirror.

[0132] Step 810: Select the frequency point set in ampCur that is greater than ceilSmallAmp and obtain iBinNoSmall.

[0133] Step 811: Perform weighted averaging of ampCur(iBinNoSmall) on diffMirror(iBinNoSmall) to obtain slopeNoSmallCur.

[0134] Step 812: Perform historical smoothing on slopeNoSmallCur to obtain slopeNoSmallSmth.

[0135] Step 813: Perform a weighted average of slopeBigSmth and SlopeNoSmallSmth to obtain slopeAvg, and then calculate angleCur based on the latter.

[0136] Step 814: Determine TDOA based on angleCur.

[0137] The TDOA estimation method of the embodiment of the present application includes the following effects:

[0138] 1. Eliminates the limitation of sampling period on estimation resolution and improves estimation accuracy. Figure 9 .exist Figure 9 a to Figure 9 In c, distMic is 0.2 meters (m), fs is 16 kilohertz (kHz), the frame length lenFrame is 512, the Fourier transform length lenFft is 2048, the signal-noise ratio (SNR) is 30 decibels (dB), and the incident angle is -20 degrees.

[0139] Figure 9 (a) shows the amplitude spectrum of the lower mic; Figure 9 (b) shows the amplitude spectrum of the upper microphone; Figure 9 (c) shows angleCur.

[0140] According to the analysis of this example, the estimated value of angleCur is very close to the actual incident angle and is not limited by the sampling period.

[0141] Second, there is no need to perform large-scale matrix search, which significantly reduces the computational complexity.

[0142] From the perspective of each frame processing, there are only a small amount of arithmetic operations and a very small amount of history smoothing, which shows that the computational complexity of this method is very low.

[0143] 3. It has good noise resistance and strong engineering practicality.

[0144] For easier understanding, see Figure 10 .exist Figure 10In the example, distMic is 0.2m, fs is 16kHz, frame length lenFrame is 512, Fourier transform length lenFft is 2048, signal-to-noise ratio SNR is 0dB, and the incident angle is -20 degrees.

[0145] Figure 10 (a) shows the amplitude spectrum of the lower mic; Figure 10 (a) shows the amplitude spectrum of the upper microphone; Figure 10 (a) shows angleCur.

[0146] According to the analysis of this example, even in scenarios with very low snr, the average error of angleCur estimation is about 5 degrees, which is still of high reference value in engineering.

[0147] It should be noted that the TDOA estimation method provided in the embodiments of the present application can be executed by a TDOA estimation device, or by a control module in the TDOA estimation device for executing the TDOA estimation method. In the embodiments of the present application, the TDOA estimation device provided in the embodiments of the present application is described by taking the execution of the TDOA estimation method by the TDOA estimation device as an example.

[0148] See also Figure 11 , Figure 11 It is a structural diagram of the TDOA estimation device provided in an embodiment of the present application.

[0149] like Figure 11 As shown, the TDOA estimation device 1100 includes:

[0150] The first acquisition module 1101 is configured to acquire the amplitude spectrum and the unwarped phase difference spectrum of the i-th speech frame, where i is a positive integer;

[0151] A second acquisition module 1102 is configured to acquire a frequency point located to the left of the mirror frequency point in the amplitude spectrum to obtain a first frequency point set of the i-th speech frame;

[0152] A first determining module 1103 is configured to determine a first phase difference slope spectrum corresponding to the i-th speech frame based on a dewarped phase difference set corresponding to the first frequency point set in the dewarped phase difference spectrum and a first amplitude set corresponding to the first frequency point set in the amplitude spectrum;

[0153] The second determining module 1104 is configured to determine the TDOA of the i-th speech frame according to a first smoothed phase difference frequency spectrum corresponding to the first phase difference frequency spectrum.

[0154] In some embodiments, the second determining module includes:

[0155] a first determining submodule, configured to determine, based on a first smoothed phase difference slope spectrum corresponding to the first phase difference slope spectrum, frequency points at which mirror warping occurs in the dewarped phase spectrum, to obtain a second frequency point set for the i-th speech frame;

[0156] a de-imaging submodule, configured to de-imaging the dewarped phase difference spectrum according to the second frequency point set to obtain a de-imaging phase difference spectrum;

[0157] A second determining submodule is configured to determine a third frequency point set of the i-th speech frame according to the amplitude mean corresponding to the amplitude spectrum, wherein the amplitude value corresponding to each frequency point in the third frequency point set is greater than the amplitude mean;

[0158] a third determining submodule, configured to determine a second phase difference slope spectrum corresponding to the i-th speech frame based on the de-mirroring phase difference set corresponding to the third frequency point set in the de-mirroring phase difference spectrum and the second amplitude set corresponding to the first frequency point set in the amplitude spectrum;

[0159] The fourth determining submodule is configured to determine the TDOA of the i-th voice frame according to the first smoothed phase difference slope spectrum and a second smoothed phase difference slope spectrum corresponding to the second phase difference slope spectrum.

[0160] In some embodiments, the amplitude spectrum is obtained by performing Fourier transform on the i-th speech frame;

[0161] The first determining submodule includes:

[0162] a first determining unit, configured to determine a first frequency point according to a maximum slope value in a first smoothed phase difference slope spectrum corresponding to the first phase difference slope spectrum and a Fourier transform length of the Fourier transform;

[0163] The second determining unit is configured to determine, based on the first frequency point and the maximum slope value, a frequency point at which mirror warping occurs, to obtain a second frequency point set of the i-th voice frame.

[0164] In some embodiments, the fourth determining submodule includes:

[0165] a third determining unit, configured to determine an average phase difference slope spectrum according to the first smoothed phase difference slope spectrum and a second smoothed phase difference slope spectrum corresponding to the second phase difference slope spectrum;

[0166] a fourth determining unit, configured to determine an angle spectrum corresponding to an i-th speech frame according to the average phase difference slope spectrum;

[0167] The fifth determining unit is configured to determine the TDOA of the i-th speech frame according to the angle spectrum.

[0168] In some embodiments, the first acquisition module includes:

[0169] a fifth determining submodule, configured to determine a threshold value corresponding to the i-th speech frame according to the amplitude mean and the amplitude maximum of the amplitude spectrum;

[0170] A first smoothing submodule is configured to perform historical smoothing on the threshold value to obtain an amplitude threshold value corresponding to the amplitude spectrum;

[0171] an acquisition submodule, configured to acquire frequency points in the amplitude spectrum that meet a first condition, and obtain a first frequency point set of the i-th speech frame;

[0172] The first condition includes: being located on the left side of the mirror frequency point, and the corresponding amplitude value being greater than the amplitude threshold value.

[0173] In some embodiments, the i-th speech frame is collected by a first microphone and a second microphone, and the amplitude spectrum is obtained by performing Fourier transform on the i-th speech frame;

[0174] The device further comprises:

[0175] The third determination module is configured to determine the mirror frequency based on the distance between the first microphone and the second microphone, the sampling frequency of the first microphone, and the Fourier transform length of the Fourier transform. The TDOA estimation device in the embodiment of the present application can be a device, or a component, integrated circuit, or chip in a terminal. The device can be a mobile electronic device or a non-mobile electronic device. Exemplarily, the mobile electronic device can be a mobile phone, tablet computer, laptop computer, PDA, in-vehicle electronic device, wearable device, ultra-mobile personal computer (UMPC), netbook or personal digital assistant (PDA), etc. The non-mobile electronic device can be a server, network attached storage (NAS), personal computer (PC), television (TV), ATM or self-service machine, etc., which is not specifically limited in the embodiment of the present application.

[0176] The TDOA estimation device in the embodiment of the present application may be a device having an operating system. The operating system may be an Android operating system, an iOS operating system, or other possible operating systems, which are not specifically limited in the embodiment of the present application.

[0177] The TDOA estimation device provided in the embodiment of the present application can achieve Figure 1 To avoid repetition, the various processes implemented by the TDOA estimation device in the method embodiment will not be described here.

[0178] Alternatively, as Figure 12 As shown, an embodiment of the present application also provides an electronic device 1200, including a processor 1201, a memory 1202, and a program or instruction stored in the memory 1202 and executable on the processor 1201. When the program or instruction is executed by the processor 1201, each process of the above-mentioned TDOA estimation method embodiment is implemented, and the same technical effect can be achieved. To avoid repetition, it will not be described here.

[0179] It should be noted that the electronic devices in the embodiments of the present application include the mobile electronic devices and non-mobile electronic devices mentioned above.

[0180] Figure 13 A schematic diagram of the hardware structure of an electronic device implementing an embodiment of the present application.

[0181] The electronic device 1300 includes but is not limited to: a radio frequency unit 1301, a network module 1302, an audio output unit 1303, an input unit 1304, a sensor 1305, a display unit 1306, a user input unit 1307, an interface unit 1308, a memory 1309, and a processor 13010 and other components.

[0182] Those skilled in the art will understand that the electronic device 1300 may also include a power source (such as a battery) to power each component, and the power source may be logically connected to the processor 1130 through a power management system, thereby implementing functions such as charging, discharging, and power consumption management through the power management system. Figure 13 The electronic device structure shown in the figure does not constitute a limitation on the electronic device. The electronic device may include more or fewer components than shown in the figure, or combine certain components, or arrange the components differently, which will not be repeated here.

[0183] The processor 13010 is configured to:

[0184] Get the amplitude spectrum and unwrapped phase difference spectrum of the i-th speech frame, where i is a positive integer;

[0185] Obtaining a frequency point located to the left of the mirror frequency point in the amplitude spectrum to obtain a first frequency point set of the i-th speech frame;

[0186] Determining a first phase difference slope spectrum corresponding to the i-th speech frame according to the unwarped phase difference set corresponding to the first frequency point set in the unwarped phase difference spectrum and the first amplitude set corresponding to the first frequency point set in the amplitude spectrum;

[0187] The TDOA of the i-th speech frame is determined according to a first smoothed phase difference frequency spectrum corresponding to the first phase difference frequency spectrum.

[0188] It should be noted that the electronic device 1300 can achieve Figure 1 To avoid repetition, the various processes implemented by the TDOA estimation device in the method embodiment will not be described here.

[0189] It should be understood that in an embodiment of the present application, the input unit 1304 may include a graphics processing unit (GPU) 13041 and a microphone 13042, and the graphics processor 13041 processes the image data of a static picture or video obtained by an image capture device (such as a camera) in a video capture mode or an image capture mode. The display unit 1306 may include a display panel 13061, and the display panel 13061 may be configured in the form of a liquid crystal display, an organic light emitting diode, etc. The user input unit 1307 includes a touch panel 13071 and other input devices 13072. The touch panel 13071 is also called a touch screen. The touch panel 13071 may include two parts: a touch detection device and a touch controller. Other input devices 13072 may include but are not limited to a physical keyboard, function keys (such as volume control keys, switch keys, etc.), a trackball, a mouse, and an operating stick, which will not be repeated here. The memory 1309 can be used to store software programs and various data, including but not limited to applications and operating systems. The processor 13010 may integrate an application processor and a modem processor, wherein the application processor primarily processes the operating system, user interface, and application programs, and the modem processor primarily processes wireless communications. It is understood that the modem processor may not be integrated into the processor 13010.

[0190] An embodiment of the present application also provides a readable storage medium, on which a program or instruction is stored. When the program or instruction is executed by a processor, the various processes of the above-mentioned TDOA estimation method embodiment are implemented and the same technical effect can be achieved. To avoid repetition, it will not be repeated here.

[0191] The processor is the processor in the electronic device described in the above embodiment. The readable storage medium includes a computer-readable storage medium, such as a computer read-only memory (ROM), random access memory (RAM), a magnetic disk, or an optical disk.

[0192] An embodiment of the present application further provides a chip, which includes a processor and a communication interface, wherein the communication interface is coupled to the processor, and the processor is used to run programs or instructions to implement the various processes of the above-mentioned TDOA estimation method embodiment, and can achieve the same technical effect. To avoid repetition, it will not be repeated.

[0193] It should be understood that the chip mentioned in the embodiments of the present application can also be called a system-level chip, a system chip, a chip system or a system-on-chip chip, etc.

[0194] It should be noted that, in this article, the terms "comprise", "include" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the statement "comprises a ..." does not exclude the presence of other identical elements in the process, method, article or device comprising the element. In addition, it should be noted that the scope of the methods and devices in the embodiments of the present application is not limited to performing functions in the order shown or discussed, and may also include performing functions in a substantially simultaneous manner or in the opposite order according to the functions involved. For example, the described method may be performed in an order different from that described, and various steps may also be added, omitted, or combined. In addition, the features described with reference to certain examples may be combined in other examples.

[0195] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus the necessary general hardware platform, and of course can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes a number of instructions for enabling a terminal (which can be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in each embodiment of the present application.

[0196] The embodiments of the present application are described above in conjunction with the accompanying drawings, but the present application is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of this application, ordinary technicians in this field can also make many forms without departing from the purpose of this application and the scope of protection of the claims, all of which are within the protection of this application.

Claims

1. A method for estimating time difference of arrival (TDOA), characterized in that: include: Obtaining an amplitude spectrum and an unwrap phase difference spectrum of an i-th speech frame, where i is a positive integer, and the unwrap phase difference spectrum is obtained by performing an unwrap operation on the phase difference spectrum of the i-th speech frame; Obtaining a frequency point located to the left of the mirror frequency point in the amplitude spectrum to obtain a first frequency point set of the i-th speech frame; performing weighted averaging on the unwarped phase difference set corresponding to the first frequency point set in the unwarped phase difference spectrum and the first amplitude set corresponding to the first frequency point set in the amplitude spectrum to obtain a first phase difference slope spectrum corresponding to the i-th speech frame; The TDOA of the i-th voice frame is determined according to the first smoothed phase difference frequency spectrum corresponding to the first phase difference slope spectrum and a mapping relationship between the smoothed phase difference slope spectrum and the TDOA of the voice frame.

2. The method according to claim 1, characterized in that The determining the TDOA of the i-th voice frame according to a first smoothed phase difference slope spectrum corresponding to the first phase difference slope spectrum includes: determining, according to a first smoothed phase difference slope spectrum corresponding to the first phase difference slope spectrum, frequency points at which mirror warping occurs in the dewarped phase difference spectrum, to obtain a second frequency point set of the i-th speech frame; de-imaging the dewarped phase difference spectrum according to the second frequency point set to obtain a de-imaging phase difference spectrum; determining, according to the amplitude mean corresponding to the amplitude spectrum, a third frequency point set of the i-th speech frame, wherein the amplitude value corresponding to each frequency point in the third frequency point set is greater than the amplitude mean; Determining a second phase difference slope spectrum corresponding to the i-th speech frame according to a de-imaging phase difference set corresponding to the third frequency point set in the de-imaging phase difference spectrum and a second amplitude set corresponding to the first frequency point set in the amplitude spectrum; The TDOA of the i-th voice frame is determined according to the first smoothed phase difference slope spectrum and a second smoothed phase difference slope spectrum corresponding to the second phase difference slope spectrum.

3. The method according to claim 2, characterized in that The amplitude spectrum is obtained by performing Fourier transform on the i-th speech frame; The determining, based on a first smoothed phase difference slope spectrum corresponding to the first phase difference slope spectrum, frequency points at which mirror warping occurs in the dewarped phase difference spectrum to obtain a second frequency point set for the i-th voice frame includes: determining a first frequency point according to a maximum slope value in a first smoothed phase difference slope spectrum corresponding to the first phase difference slope spectrum and a Fourier transform length of the Fourier transform; The frequency point where the mirror wrapping occurs is determined according to the first frequency point and the maximum slope value, to obtain a second frequency point set of the i-th voice frame.

4. The method according to claim 2, characterized in that The determining, according to the first smoothed phase difference slope spectrum and a second smoothed phase difference slope spectrum corresponding to the second phase difference slope spectrum, of the TDOA corresponding to the i-th voice frame includes: determining an average phase difference slope spectrum according to the first smoothed phase difference slope spectrum and a second smoothed phase difference slope spectrum corresponding to the second phase difference slope spectrum; Determining an angle spectrum corresponding to the i-th speech frame according to the average phase difference slope spectrum; The TDOA of the i-th speech frame is determined according to the angle spectrum.

5. The method according to claim 1, wherein The acquiring of the frequency points located to the left of the mirror frequency point in the amplitude spectrum to obtain the first frequency point set of the i-th speech frame includes: Determining a threshold value corresponding to the i-th speech frame according to the amplitude mean and the amplitude maximum of the amplitude spectrum; Performing historical smoothing on the threshold value to obtain an amplitude threshold value corresponding to the amplitude spectrum; Obtaining frequency points that meet a first condition in the amplitude spectrum to obtain a first frequency point set of the i-th speech frame; The first condition includes: being located on the left side of the mirror frequency point, and the corresponding amplitude value being greater than the amplitude threshold value.

6. The method according to claim 1, characterized in that The i-th speech frame is collected by a first microphone and a second microphone, and the amplitude spectrum is obtained by performing Fourier transform on the i-th speech frame; Before obtaining the frequency points located to the left of the mirror frequency points in the amplitude spectrum to obtain the first frequency point set of the i-th speech frame, the method further includes: The mirror frequency point is determined according to a distance between the first microphone and the second microphone, a sampling frequency of the first microphone, and a Fourier transform length of the Fourier transform.

7. A TDOA estimation device, characterized in that: include: A first acquisition module is configured to acquire an amplitude spectrum and an unwrap phase difference spectrum of an i-th speech frame, where i is a positive integer, and the unwrap phase difference spectrum is obtained by unwrap- ing the phase difference spectrum of the i-th speech frame; A second acquisition module is configured to acquire a frequency point located to the left of the mirror frequency point in the amplitude spectrum to obtain a first frequency point set of the i-th speech frame; a first determining module, configured to perform weighted averaging on a dewarped phase difference set corresponding to the first frequency point set in the dewarped phase difference spectrum and a first amplitude set corresponding to the first frequency point set in the amplitude spectrum to determine a first phase difference slope spectrum corresponding to an i-th speech frame; The second determination module is configured to determine the TDOA of the i-th voice frame according to the first smoothed phase difference frequency spectrum corresponding to the first phase difference slope spectrum and a mapping relationship between the smoothed phase difference slope spectrum and the TDOA of the voice frame.

8. The device according to claim 7, characterized in that The first acquisition module includes: a fifth determining submodule, configured to determine a threshold value corresponding to the i-th speech frame according to the amplitude mean and the amplitude maximum of the amplitude spectrum; A first smoothing submodule is configured to perform historical smoothing on the threshold value to obtain an amplitude threshold value corresponding to the amplitude spectrum; an acquisition submodule, configured to acquire frequency points in the amplitude spectrum that meet a first condition, and obtain a first frequency point set of the i-th speech frame; The first condition includes: being located on the left side of the mirror frequency point, and the corresponding amplitude value being greater than the amplitude threshold value.

9. An electronic device, characterized in that: It includes a processor, a memory, and a program or instruction stored in the memory and executable on the processor, wherein the program or instruction, when executed by the processor, implements the steps of the TDOA estimation method as described in any one of claims 1 to 6.

10. A readable storage medium, characterized in that: The readable storage medium stores a program or instruction, and when the program or instruction is executed by the processor, the steps of the TDOA estimation method as described in any one of claims 1 to 6 are implemented.

Images