Beam forming positioning method of microphone array suitable for mirror image sound source positioning correction

Abstract

The invention discloses a beam forming positioning method of a microphone array suitable for mirror image sound source positioning correction, which relates to the field of sound source positioning, and comprises the following steps of: 1, counteracting the influence of a coherent source through a reflection eliminator, and bringing the reflection eliminator into a steering vector; 2, a beam forming algorithm is optimized, white noise gain is introduced, controlled minimization of the influence of a source is achieved, and a mirror image source influence minimization technology is adopted for the controlled minimization; 3, performing simulation test; compared with a method of superposing the steering vector of the mirror image sound source and the original steering vector by a reflection eliminator method, the method provided by the invention reduces the output of the mirror image sound source direction by adopting additional constraint conditions under the condition that the original steering vector is kept unchanged, and adopts a control mechanism to keep robustness. Meanwhile, the method is expanded to the situation that multiple mirror image sources exist, and the superiority of the method is confirmed through a simulation test.

Description

Technical Field

[0001] The present invention relates to the technical field of sound source positioning, and in particular to a beamforming positioning method of a microphone array suitable for image sound source positioning correction. Background Art

[0002] Sound source localization technology is widely used in many fields such as military, aviation, aerospace, communications, transportation, and medicine. For example, it can detect the location of explosion points or gun firing points, detect submarines with sonar, and detect and image various underwater terrains and relics with acoustic devices. It can also study the distribution of noise sources when aerospace vehicles are conducting aerodynamic tests in wind tunnels, and it can locate and capture vehicles that violate the regulations by honking their horns in traffic control systems.

[0003] Generally speaking, the sound source localization algorithms based on microphone arrays are divided into three categories: one is the method based on beamforming; the second is the method based on high-resolution spectrum estimation; and the third is the method based on time delay difference of arrival (TDOA). Among them, beamforming is a controllable beamforming technology based on maximum output power. Its basic idea is to form a beam by weighted summing of the signals collected by each array element, and guide the beam by searching for the possible location of the sound source, and modify the weights to maximize the output signal power of the microphone array. That is, by adjusting the phase of the sound wave signal received by each microphone, so that it is coherently superimposed in a specific direction, thereby enhancing the sound wave signal strength in that direction. By adjusting the signal phase of each microphone, the sound wave signal strength in a specific direction is maximized, thereby determining the location of the sound source.

[0004] This method can be used in both the time domain and the frequency domain. Its time translation in the time domain is equivalent to the phase delay in the frequency domain. In the frequency domain processing, a matrix containing the autospectrum and the cross-spectrum is first used, which is called the cross-spectral matrix (CSM). At each frequency of interest, the processing of the array signal gives the energy level at each given spatial scanning grid point or each signal arrival direction (Direction of Arrival, DOA). Therefore, the array represents a summed response quantity associated with the distribution of the sound source. This method is suitable for large microphone arrays, and because of its strong adaptability to the test environment, it has also become the current mainstream acoustic positioning method.

[0005] Based on high-resolution spectral estimation

[0006] Methods based on high-resolution spectrum estimation include autoregressive AR model, minimum variance spectrum estimation (MV) and eigenvalue decomposition method (such as Music algorithm), etc. All these methods calculate the correlation matrix of spatial spectrum by acquiring the signal of microphone array. In theory, the direction of sound source can be effectively estimated. In practice, if a better accuracy is to be obtained, a large amount of calculation is required and more assumptions are needed. When the array is large, the amount of calculation of this spectrum estimation method is large, it is sensitive to environmental noise, and it is easy to cause inaccurate positioning. Therefore, it is rarely used in modern large-scale sound source positioning systems.

[0007] Time Difference of Arrival (TDOA)

[0008] This method locates the sound source by estimating the time difference of arrival (TDOA) of the sound between different microphones in the microphone array. TDOA estimation can be achieved through a variety of methods, including cross-correlation method, generalized cross-correlation method, adaptive filter method, cross-power spectrum phase method and high-order statistics method. Once TDOA is obtained, the time difference information can be combined with the spatial geometric relationship of the microphone array to perform azimuth estimation to determine the position of the sound source. The main methods include angular distance positioning method, spherical interpolation method, linear interpolation method and objective function space search positioning method.

[0009] Existing beamforming algorithms are mainly based on open test environments. The key technical point is to identify specific sound sources from the test background interference and accurately locate them. Since there are usually significant spectral differences between background interference and target sound sources, this can be used to identify them.

[0010] The form of the steering vector and spectrum is the key to beamforming. Therefore, many optimizations based on beamforming algorithms are to optimize the spectrum form (self-spectrum, cross-spectrum, de-diagonal cross-spectrum, principal component, etc.) or steering vector (theoretical value, simulation calculation value, measured value, etc.). Under the same conditions such as spectrum form, it is obvious that the closer the steering column vector is to the actual situation in the actual test environment, the more accurate the positioning result. The physical meaning of each element of the steering vector is very clear. Its exponential term represents the delay of the sound wave emitted by a grid point on the sound source received by the mth microphone; it represents the distance normalization information. Corresponding to different normalization methods, different distance normalization factors will be multiplied here. However, in fact, the steering column vectors in general beamforming algorithms are all theoretical values ​​under ideal free field conditions, because each element of the steering column vector only contains the delay corresponding to the direct sound propagation distance. When there are one or more rigid reflection surfaces in the actual test environment, how to solve or even whether these reflection information can be used, the key is to find a steering column vector that can reflect the actual sound reflection conditions.

[0011] At present, the methods of using microphone arrays to locate sound sources can be generally divided into two categories: directional wave velocity formation and time delay estimation. Directional wave velocity formation is a form of the maximum output power controllable beam method. This method filters and delays the speech signal received by the microphone array and performs weighted summation to form a beam with a specific direction, and then calculates the beam energy in each direction. The direction with the highest power in the search space is the maximum approximation of the sound source position. In the method based on time delay estimation, since there is a time difference between the signal emitted by the sound source reaching each microphone, this method estimates this time delay based on the correlation between the signals of each channel, thereby estimating the sound source position.

[0012] The method of directional wave velocity formation needs to scan the entire space in order to search for the position that can maximize the beam power, which requires a lot of calculations and is therefore difficult to meet real-time requirements.

[0013] Methods based on time delay estimation have fast calculation speed and can meet real-time requirements. Such methods can be applied to non-stationary broadband sources such as speech, but they work better when there is only a single source. The accuracy of positioning is related to the estimation accuracy of the time difference of arrival (TDOA).

[0014] In summary, it is difficult to improve the performance of sound source identification when there are one or more rigid reflective surfaces in the test environment when locating the sound source in the prior art. To this end, a beamforming positioning method of a microphone array suitable for image sound source positioning correction is provided to solve the above problem. Summary of the invention

[0015] In view of the problems existing in the above-mentioned prior art, the present invention provides a beamforming positioning method for microphone arrays suitable for image sound source positioning correction. When there are one or more rigid reflective surfaces in the test environment, a method for improving the performance (spatial resolution, dynamic range, etc.) of sound source identification by using reflected wave information is provided based on beamforming. First, the influence of the coherent (image) source is offset by a "reflection eliminator", and on this basis, the beamforming algorithm is further optimized to achieve "controlled minimization" of the influence of the (image) source. This method can effectively distinguish reflected sound sources and realize sound source positioning.

[0016] In order to achieve the above object, the microphone array used in the present invention is suitable for a beamforming positioning method for image sound source positioning correction, including:

[0017] Step 1: Use a reflection canceller to cancel the influence of the coherent source and incorporate the reflection canceller into the steering vector;

[0018] Step 2: Optimize the beamforming algorithm, introduce white noise gain, and achieve "controlled minimization" of the source's influence. "Controlled minimization" uses the technology of minimizing the influence of the image source.

[0019] Step 3: Simulate the test to simulate the environment inside the switch cabinet: Assume that there is no wind, the microphone array is at a medium height above the ground, the sound source is located at (0,0,0.8), the microphone array is 3 meters away from the front reflection wall horizontally, and the reflection wall is 2.22 meters high; observe the difference in the sound source and mirror sound source effect diagrams formed by the traditional beamforming method and the beamforming positioning method.

[0020] As a further optimization of the above scheme, the reflection eliminator is used to reduce the noise of the sound source. The reflection eliminator is based on the principle of wave mechanics. A shielding plate is set between the sound source and the muffler so that the energy of the sound wave is reflected. The sound wave reflected by the shielding plate interferes with the original sound wave, thereby eliminating part of the noise.

[0021] As a further optimization of the above scheme, the white noise gain refers to an indicator for measuring the robustness of the microphone array to the microphone self-noise in microphone array signal processing, and is used to evaluate the performance of the array in a white noise environment;

[0022] White noise refers to noise whose power spectrum density is uniformly distributed throughout the frequency domain. Random noise with the same energy density at all frequencies is called white noise.

[0023] The microphone array of the present invention is applicable to the beamforming positioning method for image sound source positioning correction, and has the following beneficial effects:

[0024] The microphone array of the present invention is suitable for the beamforming positioning method of the image sound source positioning correction. Compared with the method of superimposing the steering vector of the image sound source with the original steering vector by the reflection eliminator method, the method proposed by the present invention reduces the output in the direction of the image sound source by adopting additional constraints while keeping the original steering vector unchanged, and adopts a control mechanism to maintain robustness. At the same time, the method is extended to the situation where there are multiple image sources, and the simulation test confirms the superiority of the method.

[0025] With reference to the following description and drawings, a specific embodiment of the present invention is disclosed in detail, indicating the manner in which the principles of the present invention can be adopted. It should be understood that the scope of the embodiments of the present invention is not limited thereby, and within the spirit and scope of the appended claims, the embodiments of the present invention include many changes, modifications and equivalents. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 A schematic diagram of the simulated switch cabinet environment of the present invention;

[0027] Figure 2 The imaging effect of the sound source at (0,0,0.8) and the mirror image sound source at (0,0,1.42) using the traditional method of the present invention is shown in FIG. Figure 1 ;

[0028] Figure 3 This is a diagram showing the imaging effect of the sound source at (0,0,0.8) and the mirror image sound source at (0,0,1.42) by using mirror image minimization of the present invention;

[0029] Figure 4 A schematic diagram of the relative position distribution of the sound source and the image sound source of the present invention;

[0030] Figure 5 The imaging effect of the sound source at (0,0,0.8) and the mirror image sound source at (0,0,1.42) using the traditional method of the present invention is shown in FIG. Figure 2 ;

[0031] Figure 6 This is a diagram of the imaging effect of using the reflection eliminator of the present invention on the sound source at (0,0,0.8) and the mirror image sound source at (0,0,1.42). DETAILED DESCRIPTION

[0032] Please refer to the instruction manual Figure 1-6 The present invention provides a technical solution: a microphone array is suitable for a beamforming positioning method for image sound source positioning correction.

[0033] The application scenario of the present invention is a closed test area. The main difference from an open test environment is that one or more rigid reflection surfaces will be formed in a limited closed space. When the aerodynamic sound source is close to the reflection wall, the result of the traditional beamforming algorithm will be interfered by the nearby image sound source, and the image sound source is coherent with the original sound source. The recalculated sound source spectrum often deviates from the spectrum of the same sound source obtained in an open environment, and the spectrum will show periodic modulation. This phenomenon is particularly obvious at low frequencies. Based on this, the present invention studies an unconventional beamforming technology. First, the influence of the coherent (image) source is offset by a "reflection eliminator". On this basis, the beamforming algorithm is further optimized to achieve "controlled minimization" of the influence of the (image) source.

[0034] First, the technique of adding mirror sources to the transmission model is discussed. It is found that this technique is not very applicable because it lacks robustness.

[0035] Then, a more robust beamforming technique was proposed that minimizes the effect of a given image source. With this technique, the results were much better. However, this approach still lacked robustness at low frequencies.

[0036] Therefore, a modification to the beamforming technique is proposed to maintain its robustness. Using the “controlled minimization” technique, the best agreement was found between the recomputed spectra of the source close to the wall and the same source under anechoic conditions.

[0037] The purpose of the present invention is to provide a method for improving the performance (spatial resolution, dynamic range, etc.) of sound source identification by using reflected wave information based on beamforming when there are one or more rigid reflective surfaces in the test environment. First, the influence of the coherent (image) source is offset by a "reflection eliminator", and on this basis, the beamforming algorithm is further optimized to achieve "controlled minimization" of the influence of the (image) source. This method can effectively distinguish the reflected sound source and realize the localization of the sound source.

[0038] However, traditional beamforming technology is only suitable for open measurement scenarios with wind speed interference.

[0039] Simulation test environment:

[0040] As attached Figure 1 As shown in , the environment inside the switch cabinet is simulated: assuming that there is no wind, the microphone array is at a medium height above the ground, the sound source is located at (0,0,0.8), which is 0.8 meters above the coordinate origin, the microphone array is 3 meters horizontally away from the front reflective wall, and the height of the reflective wall is 2.22 meters.

[0041] As attached Figure 2 The traditional method is used to image the sound source at (0,0,0.8) and the mirror image sound source at (0,0,1.42). Figure 5 The imaging effect of the sound source at (0,0,0.8) and the mirror sound source at (0,0,1.42) is shown in the figure using mirror minimization.

[0042] As attached Figure 3 The traditional method is used to image the sound source at (0,0,0.8) and the mirror image sound source at (0,0,1.42). Figure 6 The middle one shows the imaging effect of the sound source at (0,0,0.8) and the mirror image sound source at (0,0,1.42) using a reflection eliminator. It can be seen from the figure that Figure 3 The sound source position coincides with the peak position only when the frequency is higher than 2000Hz. Figure 6 The sensitivity to the assumed source location is much higher in .

[0043] The purpose of the reflection eliminator in the present invention is to offset the influence of the coherent (image) source and incorporate it into the steering vector. The reflection eliminator is used to reduce the noise of the sound source. The reflection eliminator is based on the principle of wave mechanics. By setting a shielding plate between the sound source and the muffler, the energy of the sound wave is reflected. The sound wave reflected by the shielding plate interferes with the original sound wave, eliminating part of the noise.

[0044] Reflection eliminators are suitable for scenarios where local noise reduction is required, such as installing reflection eliminators in certain specific areas or equipment to reduce the impact of noise on the surrounding environment. Due to its simple design and low cost, reflection eliminators have certain applications in industrial and civil fields.

[0045] In general, reflection eliminators achieve noise reduction through the principles of reflection and interference, and are suitable for scenarios that require simple, low-cost noise reduction.

[0046] The "controlled minimization" in the present invention adopts the image source influence minimization technology, and the image source influence minimization technology is highly sensitive to the assumed sound source position.

[0047] In order to control the robustness of "controlled minimization", "white noise gain" is introduced. "White noise gain" refers to an indicator that measures the robustness of the microphone array to the self-noise of the microphone in microphone array signal processing. White noise gain can be regarded as a special case of signal-to-noise ratio gain, which is used to evaluate the performance of the array in a white noise environment.

[0048] Under the corresponding assumptions, the white noise gain can be defined as the norm of the beamformer coefficients. The white noise gain is very important in microphone array signal processing because it provides a measure of the robustness of the microphone's self-noise. In practical applications, increasing the number of array elements can significantly increase the white noise gain, thereby enhancing the performance of the array in a white noise environment; white noise refers to noise whose power spectral density is uniformly distributed throughout the frequency domain. Random noise with the same energy density at all frequencies is called white noise. In the quantization process, since the signal is quantized to a value on a fixed grid, errors will occur. This error appears as white noise, which is a prior art and will not be elaborated here.

[0049] Compared with the reflection eliminator method that superimposes the steering vector of the image sound source with the original steering vector, the method proposed by the present invention reduces the output in the direction of the image sound source by adopting additional constraints while keeping the original steering vector unchanged, and adopts a control mechanism to maintain robustness. At the same time, the method is extended to the situation where there are multiple image sources, and simulation experiments confirm the superiority of the method.

Claims

1. A microphone array is suitable for a beamforming positioning method for image sound source positioning correction, characterized in that: include: Step 1: Use a reflection canceller to cancel the influence of the coherent source and incorporate the reflection canceller into the steering vector; Step 2: Optimize the beamforming algorithm, introduce white noise gain, and achieve "controlled minimization" of the source's influence. "Controlled minimization" uses the technology of minimizing the influence of the image source. Step 3: Simulate the test to simulate the environment inside the switch cabinet: Assume that there is no wind, the microphone array is at a medium height above the ground, the sound source is located at (0,0,0.8), the microphone array is 3 meters away from the front reflection wall horizontally, and the reflection wall is 2.22 meters high; observe the difference in the sound source and mirror sound source effect diagrams formed by the traditional beamforming method and the beamforming positioning method.

2. The microphone array according to claim 1 is suitable for a beamforming positioning method for image sound source positioning correction, characterized in that: The reflection eliminator is used to reduce the noise of the sound source. The reflection eliminator is based on the principle of wave mechanics. By setting a shielding plate between the sound source and the silencer, the energy of the sound wave is reflected. The sound wave reflected by the shielding plate interferes with the original sound wave, thereby eliminating part of the noise.

3. The microphone array according to claim 1 is suitable for a beamforming positioning method for image sound source positioning correction, characterized in that: The white noise gain refers to an indicator for measuring the robustness of a microphone array to microphone self-noise in microphone array signal processing, and is used to evaluate the performance of the array in a white noise environment; White noise refers to noise whose power spectral density is uniformly distributed throughout the frequency domain. Random noise with the same energy density at all frequencies is called white noise.

Images