A localization method for several known possible sound sources in a locally complex stereo environment

Abstract

The present invention relates to the technical field of sound source localization, and specifically to a method for localizing several known possible sound sources in a locally complex stereo environment. The method comprises preparing a plurality of active sound sources for position calibration, and placing the active sound source approximately above a potential sound source. A microphone in a microphone array is set as a reference microphone, and the relationship between the active sound source and the reference microphone and each microphone is calculated through the cross-correlation of signals. The method has the advantages of reducing the number of sound sources and finding the component most likely to cause a sound fault from a limited number of sound sources in a stereo environment. Compared with the traditional method of localizing the position of sound sources, the present invention aims to localize the component, and is therefore suitable for long-term abnormal sound fault monitoring in some closed and dangerous spaces with a large number of known sound sources.

Description

Technical Field

[0001] The present invention relates to the technical field of sound source localization, and in particular to a method for localizing several known possible sound sources in a locally complex stereo environment. Background Art

[0002] Currently, many laboratories are inaccessible for personnel to conduct on-site measurements and experience due to security concerns. Without array-based measurement equipment, only the amplitude at a specific point can be measured. If a microphone array is deployed, traditional acoustic cameras require significant computational effort, resulting in a massive data throughput per second, making them suitable only for short-term processing. Over extended monitoring periods, the sheer volume of data throughput makes recording difficult, requiring significant hard drive capacity and exposing the system to the risk of system failure. Summary of the Invention

[0003] The present invention aims to reduce the number of sound sources and identify the component most likely to cause a sound fault from a limited number of sound sources in a stereo environment. A method for localizing several known possible sound sources in a locally complex stereo environment is provided, characterized by the following steps:

[0004] S1. Arrange the microphone array at any position in the scene, wherein the microphone array is a three-dimensional array;

[0005] S2. Identify potential sound source locations:

[0006] Prepare multiple active sound sources for position calibration, and place the active sound sources roughly above the potential sound source;

[0007] One microphone in the microphone array is set as a reference microphone, and the relationship between the active sound source and the reference microphone and each microphone is calculated by cross-correlation of the signals;

[0008] S3. Move several sampling points back and forth to obtain the sampling point matrix of the potential sound source;

[0009] S4. Monitor the sound source in real time using the sampling point matrix of the potential sound source. Based on the sampling point matrix, shift the microphones other than the reference microphone according to the required signal. Finally, obtain multiple signal waveforms with lengths corresponding to the set sampling points. Superimpose these multiple waveforms to obtain a total waveform. Calculate weights based on the total waveform, compare the weights, and determine the location of the potential sound source.

[0010] The method of the present invention also has the following preferred technical solutions:

[0011] 1. The microphone array is spherical or polyhedral.

[0012] 2. Synchronize the AD data collector to collect the signal, and calculate the relationship between the active sound source and the reference microphone and each microphone through the cross-correlation of the signals. For discrete signals, the formula is expressed as R(n)=(1 / N)∑[x(m)y(m+n)], where n represents the nth sampling point, N represents the total number of sampling points, and m represents the sampling point variable, where m varies from 0 to N-1. Through cross-correlation calculation, we can obtain the cross-correlation curve and obtain the left or right of the position 0 where the cross-correlation signal has the maximum value.

[0013] 3. The potential sound source signal collected by the reference microphone can only basically coincide with the signal obtained by the reference microphone when the sampling points of the signals collected by other microphones are shifted a certain distance before and after.

[0014] 4. The weighting algorithm is as follows: the absolute value of the superimposed signal is calculated, that is, all the signals are converted to positive values, and then the integral of the signal is calculated, that is, each value of the signal is added up. The resulting integral value is the weight of the potential sound source at one of the positions.

[0015] Compared with the existing technology, the advantages of the present invention are: reducing the number of sound sources, and finding the component most likely to cause sound failure among a limited number of sound sources in a three-dimensional environment. Compared with the traditional positioning of the sound source, the present invention aims to locate the component, so it is suitable for long-term abnormal sound fault monitoring in some closed and dangerous spaces with a large number of known sound sources. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 is a cross-correlation function of the two microphone signals based on time in the present invention;

[0017] Figure 2 It is the cross-correlation function of the two microphone signals based on the sampling points of the present invention. DETAILED DESCRIPTION

[0018] The present invention will be further described below with reference to the accompanying drawings. The structure and principle of the present invention will be very clear to those skilled in the art. It should be understood that the specific embodiments described herein are only intended to explain the present invention and are not intended to limit the present invention.

[0019] Example: Assume that there are three possible sound sources A, B, and C in a closed space. To distinguish which sound source is emitting the sound in the space during observation, perform the following calculation steps:

[0020] 1. Install our stereo microphone array in a fixed position. This array is similar to a polyhedron structure, with high-sensitivity microphones placed at each vertex. If it is a tetrahedron, there will be five microphones making up this small array.

[0021] 2. Place an active sound source on three sound sources at a time. The active sound source can be music from a mobile phone. The signal is collected by synchronizing the AD data collector. One microphone in the stereo microphone array is set as the reference microphone. The relationship between the active sound source and the reference microphone and each microphone can be calculated through the cross-correlation of the signals. For discrete signals, the formula is expressed as R(n)=(1 / N)∑[x(m)y(m+n)], where n represents the nth sampling point, N represents the total number of sampling points, and m represents the sampling point variable, where m varies from 0 to N-1. By sampling 11,000 sampling points in each channel at a time, taking the reference microphone from 500 to 10,500 sampling points, with a length of 10,000 sampling points, and taking the second microphone also with 10,000 sampling points, the cross-correlation calculation can produce a cross-correlation curve, as shown in the following example: Figure 2 As shown. This determines whether the maximum cross-correlation signal is to the left or right of position 0, which is the number of sampling points before and after the active source between the second microphone and the reference microphone. Similarly, the relationship between sound source A and the reference microphone and the third, fourth, and fifth microphones can be determined. By shifting the sampling points back and forth, we can derive the sampling point matrix of sound source A. This means that the signal from sound source A collected by the reference microphone must be shifted a certain number of sampling points before and after the signals collected by the second to fifth microphones to essentially coincide with the signal collected by the reference microphone. The same method can be used to derive the matrix generated by sound sources B and C.

[0022] 3. After completing the matrix measurement, the sound source can be monitored in real time. The main principle is: when a sound signal is measured, in order to know where the signal is emitted from, each microphone samples 12,000 sampling points at a time, and uses the 500th to 10,500th sampling points, with a total length of 10,000 sampling points, to participate in the calculation. Assuming that the sound source is transmitted from point A, according to step 2, except for the reference microphone, the remaining microphones need to be displaced by several 17mm distances when the sound source is at point A, that is, several sampling points. According to the matrix made in step 2, the second to fifth microphones are displaced according to the required signal. Finally, five signal waveforms with a length of 10,000 sampling points are obtained. These five waveforms are superimposed to obtain a total waveform. If the sound source actually originates from point A, then the amplitude of the superimposed sound source is the largest. To remove the influence of negative signal values, the absolute value of the superimposed signal is calculated, that is, all signals are converted to positive values. The integral of the signal is then calculated by adding up each value of the signal. The resulting integral value is the weight of the sound source if it were at point A. The same method can be used to calculate the weight of the sound source at points B and C. By comparison, the one with the largest weight is the sound source.

[0023] Through a certain amount of experiments, the location of the sound source can be determined by the superposition of the sound source amplitude, but the integration method is more accurate and stable than the amplitude method.

Claims

1. A method for localizing several known possible sound sources in a local complex stereo environment, characterized by The method steps are as follows: S1. Arrange the microphone array at any position in the scene, wherein the microphone array is a three-dimensional array; S2. Identify potential sound source locations: Prepare multiple active sound sources for position calibration, and place the active sound sources roughly above the potential sound source; One microphone in the microphone array is set as a reference microphone, and the relationship between the active sound source and the reference microphone and each microphone is calculated by cross-correlation of the signals; S3. Shift the signals collected by the other microphones forward and backward by a certain number of sampling points to obtain a sampling point matrix of the potential sound source. S4. Monitor the sound source in real time through the sampling point matrix of the potential sound source, perform displacement adjustment on the signals collected by the microphones other than the reference microphone according to the sampling point matrix, and finally obtain multiple signal waveforms with lengths corresponding to the set sampling points. Superimpose the multiple waveforms to obtain a total waveform, calculate the weight of each potential sound source based on the total waveform, compare the weight of each potential sound source, and determine the position of the potential sound source.

2. A method for locating several known possible sound sources in a locally complex stereo environment as claimed in claim 1, characterized in that The microphone array is spherical or polyhedral in shape.

3. The method for locating several known possible sound sources in a locally complex stereoscopic environment as claimed in claim 1, characterized in that Synchronize the AD data collector to collect the signal, and calculate the relationship between the active sound source and the reference microphone and each microphone through the cross-correlation of the signal. The formula for discrete signals is R(n)=(1 / N)∑[x(m)y(m+n)], n represents the nth sampling point, N represents the total number of sampling points, and m represents the sampling point variable, where m varies from 0 to N-1. The cross-correlation curve can be obtained through cross-correlation calculation to determine whether it is to the left or right of the position 0 where the cross-correlation signal has the maximum value.

4. A method for locating several known possible sound sources in a locally complex stereoscopic environment as claimed in claim 1, characterized in that The potential sound source signal collected by the reference microphone requires that the sampling points of the signals collected by other microphones be shifted forward and backward by a certain amount in order to basically coincide with the signal obtained by the reference microphone.

5. The method for locating several known possible sound sources in a locally complex stereo environment as claimed in claim 1, characterized in that The weighting algorithm is as follows: the absolute value of the superimposed signal is calculated, that is, all the signals are converted to positive values, and then the integral of the signal is calculated, that is, each value of the signal is added up. The resulting integral value is the weight of the potential sound source at one of the positions.

Images