WO2023173514A1

WO2023173514A1 - Sound source positioning method under strong multipath interference condition

Info

Publication number: WO2023173514A1
Application number: PCT/CN2022/085493
Authority: WO
Inventors: 俞李斌; 王智; 林俊光; 席玉章; 孙士恩; 董益华; 马聪
Original assignee: 浙江浙能技术研究院有限公司
Priority date: 2022-03-16
Filing date: 2022-04-07
Publication date: 2023-09-21
Also published as: CN114624652A; CN114624652B

Abstract

A sound source positioning method under a strong multipath interference condition, comprising the following steps: constructing a microphone array (S1); collecting a sound source signal by means of the microphone array, thereby determining a characteristic frequency (ω _p) of the sound source signal and the sound propagation direction; calculating a signal gain ratio; selecting a reconstruction scheme of the microphone array according to the signal gain ratio, and changing the characteristic frequency (ω _p) according to the reconstruction scheme; using the characteristic frequency (ω _p) detected by the microphone array to predict the sound source direction, generating a reconstruction scheme according to the signal gain ratio in the predicted direction, reconstructing the microphone array according to the reconstruction scheme, repeating the aforementioned process to gradually reduce the influence of multipath effects on the characteristic frequency (ω _p) value detected by the microphone array while mitigating the influence on the judgment of the sound source direction, and when a change of the characteristic frequency (ω _p) before and after reconstruction is smaller than a threshold value, using a predicted direction obtained according to the characteristic frequency (ω _p) as a sound source direction.

Description

A sound source localization method under strong multipath interference conditions

Technical field

The invention relates to a sound source positioning method under strong multipath interference conditions, and belongs to the field of sound source positioning.

Background technique

Sound source localization has very important application value. For example, in a pipeline leak scenario, the leak point can be located based on the noise emitted by the leak point, so that the leak point can be discovered early and repaired. However, the sound propagation environment is complex, so the microphone array will be affected by multipath phenomena in the process of receiving sound signals, resulting in errors in determining the direction of the sound source that are difficult to eliminate.

Contents of the invention

The technical problem to be solved by the present invention is to overcome the shortcomings of the existing technology and provide a sound source positioning method under strong multipath interference conditions.

To solve the above technical problems, the present invention adopts the following technical solutions:

A sound source localization method under strong multipath interference conditions, including the following steps:

Step S1: Build a microphone array;

Step S2: The microphone array collects the sound source signal to determine the characteristic frequency ω _p of the sound source signal, p = 1, 2,..., q, and filter the sound source signal to retain the ω _p neighborhood range. The sound signal within ω p is then obtained, and the sound propagation direction corresponding to ω _p is obtained;

Step S3: Calculate the signal gain ratio of the microphone array in the sound propagation direction corresponding to ω _p ;

Step S4: Select the reconstruction scheme of the microphone array through the signal gain ratio, repeat steps S1 and S2 according to the reconstruction scheme to change ω _p , and set the threshold δ. If the change of ω _p is less than δ, then according to the changed The sound propagation direction corresponding to ω _p is used to determine the sound source position. If the change in ω _p is not less than δ, continue to repeat steps S3, S1 and S2 in sequence until the change in ω _p is less than δ.

The beneficial effects of the present invention are:

The characteristic frequency detected by the microphone array is then used to predict the direction of the sound source. A reconstruction plan is generated based on the signal gain ratio in the predicted direction, and the microphone array is reconstructed based on the reconstruction plan, and the above process is repeated. , gradually reducing the impact of the multipath effect on the numerical value of the characteristic frequency detected by the microphone array, and at the same time reducing the impact on the judgment of the direction of the sound source. Taking advantage of the positive correlation between the above two effects, when the change in the characteristic frequency before and after reconstruction is less than the threshold, It can be considered that the multipath effect can be ignored, so the predicted direction obtained based on the characteristic frequency at this time can be used as the sound source direction.

In the present invention, q is 2 or 3 or 4 or 5, and the value of q remains unchanged before and after steps S1 and S2 are repeated.

The microphone array of the present invention includes N microphones, one of which is a central microphone, and the other microphones are peripheral microphones and are arranged sequentially along the circumferential direction of the central microphone. The spherical coordinates are established with the central microphone as the origin, and any point in the spherical coordinates is connected to the central microphone. The elevation angle of the line in spherical coordinates is

And the horizontal angle is θ, the distance between this point and the j-th peripheral microphone is r _j , and the elevation angle of the line connecting this point and the j-th peripheral microphone in spherical coordinates is

And the horizontal angle is θ _j , set the function

in

v is the speed of sound, i is the imaginary unit, and the elevation angle of the sound propagation direction corresponding to ω _p in spherical coordinates is

And the horizontal angle is θ _p , then the signal gain ratio in the sound propagation direction corresponding to ω _p

When the microphone array is constructed for the first time in the present invention, the peripheral microphones are evenly distributed along the circumferential direction of the central microphone. At this time, the distance between two adjacent peripheral microphones is d _min . The position movement distance of the jth peripheral microphone before and after the microphone array is reconstructed is Δd _j . Δd _j <d _min .

The present invention uses a multi-objective optimization algorithm of quantum particle swarm to generate a reconstruction plan for the microphone array to ensure that the change in ω _p is reduced after each reconstruction of the microphone array.

In step S3 of the present invention, the signal gain ratio is calculated for the sounds of multiple characteristic frequencies in their respective propagation directions and the weighted average is calculated. In step S4, the reconstruction scheme is selected based on the weighted average of the signal gain ratio.

The present invention regards each microphone array arrangement as a kind of particle, and the distance between the current particle and the optimal particle is γ. If γ<1, then the expansion coefficient α=0.5+γ, and if γ=1, then α=1.8.

Other features and advantages of the present invention will be disclosed in detail in the following detailed description and drawings.

[Picture description]

The present invention will be further described below in conjunction with the accompanying drawings:

Figure 1 is a flow chart of a sound source localization method under strong multipath interference conditions according to Embodiment 1 of the present invention.

Detailed ways

The technical solutions of the embodiments of the present invention will be explained and described below with reference to the accompanying drawings of the embodiments of the present invention. However, the following embodiments are only preferred embodiments of the present invention and are not exhaustive. Based on the examples in the implementation mode, other embodiments obtained by those skilled in the art without any creative work shall fall within the protection scope of the present invention.

In the following description, the occurrence of terms such as "inside", "outside", "upper", "lower", "left", "right", etc. to indicate the orientation or positional relationship are only for the convenience of describing the embodiments and simplifying the description, rather than Any indication or implication that the referred device or element must have a specific orientation, be constructed and operate in a specific orientation should not be construed as a limitation on the invention.

Example 1:

The sound source emits sound with a certain characteristic frequency. Due to the influence of the multipath effect, in addition to receiving the sound directly emitted by the sound source, the microphone array will also receive a certain degree of noise. This part of the noise will interact with the sound directly emitted by the sound source. Destructive or constructive, therefore, when the microphone array performs data processing on the received sound signal, it will not only misestimate the value of the characteristic frequency at the sound source, but also misjudge the direction of the sound source. This is the main reason why multipath phenomenon interferes with sound source localization.

From this, a special phenomenon can be found, that is, the misestimation of the characteristic frequency and the misjudgment of the sound source direction occur at the same time, and the degree of the two shows a positive correlation characteristic. Based on this, this embodiment provides a strong multipath The sound source localization method under interference conditions includes the following steps:

Step S1: Construct a microphone array, in which there are a total of N microphones in the microphone array. One microphone is the central microphone, and the other microphones are peripheral microphones and are arranged sequentially along the circumferential direction of the central microphone. That is, the total number of peripheral microphones is N-1. All microphones are always in the same plane;

In the microphone array constructed for the first time, the peripheral microphones are evenly distributed along the circumferential direction of the central microphone. At this time, the distance between two adjacent peripheral microphones is d _min ;

Step S2: Set a reference direction and use the current microphone array to receive the sound signal from an unknown sound source. The microphone array can use this to determine the characteristic frequency ω _p of the sound source signal, p=1,2,... ..., q, the characteristic frequency is the frequency with greater sound intensity, so conventionally, when sorting different characteristic frequencies, the sound intensity corresponding to ω ₁ , ω ₂ ...ω _q is guaranteed Gradually decrease, under the influence of the multipath effect, the characteristic frequency value of the sound signal at the sound source and the characteristic frequency value of the sound signal received by the microphone array will be different due to the frequency offset, but the number of characteristic frequencies in the two sound signals It is still the same, that is, each characteristic frequency in the sound signal received by the microphone array can find the corresponding characteristic frequency in the sound signal at the sound source. The microphone array processes ω _p through the MUSIC algorithm, and the corresponding characteristic frequency can be obtained. The angle D _p between the sound propagation direction and the reference direction.

In order to facilitate subsequent description, the characteristic frequency obtained by the first constructed microphone array is

The characteristic frequency is

The sound signal is filtered and retained

Sound signals within the neighborhood can be calculated

The corresponding sound propagation direction, the angle between the sound propagation direction and the reference direction is

can represent

Corresponding sound propagation direction, perform the above steps multiple times on the sound signal received by the microphone array to obtain respectively

The angle between the corresponding sound propagation direction and the reference direction

Due to the multipath effect,

There is a certain difference between them. Since the sound source position is unique, therefore

The difference between them won’t be too big;

In some special environments, since sound undergoes different times of reflection when propagating to the microphone array, the

May get multiple

and different

The difference between them is large, for example if

Three values with large numerical differences between each other were obtained.

at the same time

corresponding

There is only one, then by comparison

different from

The difference between them can eliminate obvious errors

so that

and

There is a one-to-one correspondence between them;

Step S3: For one of the ω _{p s} , the signal gain ratio T _p in the corresponding sound propagation direction (referred to as the D _p direction) can be solved. The signal gain ratio T _p represents the sound intensity gradient in the D _p direction. The signal gain ratio The larger the value, the more concentrated the signal intensity is in the D _p direction. Correspondingly, the less obvious the influence of the multipath effect is in the D _p direction. In contrast, the smaller the signal gain ratio is, the more dispersed the signal intensity is in the D _p direction. The influence of multipath effect is more obvious in the D _p direction;

Step S4: Generate a reconstruction plan for the microphone array with the signal gain ratio as the optimization target, and repeat steps S1 and S2 according to the reconstruction plan to change ω _p ;

Before and after the microphone array is reconstructed, the positions of the central microphone and the reference direction will not change, and the number of microphones will not change. Only the positions of the peripheral microphones will change;

Taking the microphone array constructed for the second time as an example, the sound signal is re-collected at this time, and the current characteristic frequency is obtained:

Then perform filtering separately to obtain the angle between the corresponding sound propagation direction and the reference direction.

and

One-to-one correspondence between them,

and

There is a one-to-one correspondence between them. Since the arrangement of the microphone array changes between the first build and the second build, the corresponding

and

not equal,

and

Also not the same;

By analogy, if subsequent microphone arrays continue to be reconstructed, the characteristic frequency of the sound signal collected by the microphone array constructed for the gth time is

In the first microphone array built,

exist

The signal gain ratio in the direction is

In the second constructed microphone array,

exist

The signal gain ratio in the direction is

By analogy, in the microphone array constructed for the gth time,

exist

The signal gain ratio in the direction is

The meaning of generating the reconstruction scheme of the microphone array with the signal gain ratio as the optimization goal is to ensure that

Its meaning indicates that the multipath effect is

The influence in the direction is compared with the

is smaller in the direction, the sound signal is in

More concentrated in direction,

The direction is closer to the direction of the actual sound source. The specific generation method of the reconstruction plan will be explained later in this embodiment.

Select the sequence number p and set the threshold δ. For the microphone array constructed for the g+1th time, if

It shows that the characteristic frequency measurement error caused by the multipath effect is small enough. According to the conclusion that the characteristic frequency misestimation described in the second paragraph of this embodiment is positively correlated with the degree of misjudgment of the sound source direction,

The direction can be used as the actual sound propagation direction,

The numerical value can be used as the angle between the connection line between the sound source and the center microphone of the microphone array constructed at the g+1th time and the reference direction, thereby obtaining the sound source direction position information (ie, the sound source positioning in this embodiment).

If the change of ω _p is not less than δ, that is

Then continue to repeat steps S3, S1 and S2 in sequence, and continue to reconstruct the microphone array until the change in ω _p is less than δ, that is,

For the microphone array constructed for the gth time, the spherical coordinates are established with the center microphone as the origin. The elevation angle of any point in the spherical coordinates connected to the center microphone in the spherical coordinates is

And the horizontal angle is θ, and the distance between this point and the j-th peripheral microphone is

The elevation angle of the line connecting this point and the jth peripheral microphone in spherical coordinates is

And the horizontal angle is

set function

in

v is the speed of sound, i is the imaginary unit,

Represents the sum of the normalized signal strengths sent to the center microphone and each peripheral microphone at any point in the space. On this basis,

Corresponding sound propagation direction (abbreviation

direction) in spherical coordinates the elevation angle is

And the horizontal angle is

but

signal gain ratio in direction

Similarly, in the microphone array constructed at the g+1th time, it can be solved

Since the signal gain ratio is normalized

Calculation is performed, so even if the sound intensity at the sound source changes in real time,

and

Also comparable.

to ensure that

After the microphone array is constructed for the gth time, the multi-objective optimization algorithm of quantum particle swarm is used to generate the reconstruction plan of the microphone array.

The core idea of the multi-objective optimization algorithm of quantum particle swarm is to record each microphone array arrangement as a particle P, and the coordinates of the j-th peripheral microphone in the particle are (x _j , y _j ), then P=P (x ₁ , y ₁ , x ₂ , y ₂ ,..., x _N-1 , y _N-1 ). After the first microphone array is constructed, the particles corresponding to the microphone array are P _1, 1. At this time, M-1 particles are randomly generated, which can be recorded as P _{2, 1} , P _3, 1, .... .., P _M,1 , which can be known through the detection of particle P _1,1

and

At this time, P _1,1 , P _2,1 ,..., P _M,1 can be obtained respectively through calculation.

The signal gain ratio in the direction is T _1,1 , T _2,1 ,..., _TM,1 . (at this time

That is T _1,1 ). By calculating P _1,1 , P _2,1 ,..., PM _,1 to generate P _1,2 , P _2,2 ,..., PM _,2 , Select a particle from P _1,2 , P _2,2 ,..., _PM,2 as the reconstruction plan, so that the form of the microphone array meets the reconstruction plan when the microphone array is constructed for the second time, Then the microphone array constructed for the second time can be learned through actual inspection.

and

At this time, P _1,2 , P _2,2 ,..., P _M,2 can be obtained respectively through calculation.

The signal gain ratio in the direction is T _{1, 2} , T 2, ₂ ,..., T _M,2 , and so on. After the g-th construction of the microphone array is completed, P _{1, g} , P has been obtained _{2, g} ,..., P _M,g , at this time it can be known through detection

and

Correspondingly, through calculation we can obtain P _1,g ,P _2,g ,...,P _M,g in

The signal gain ratio in the direction is T _1,g , T _2,g ,..., T _M,g , and then calculated to obtain P _1,g+1 , P _2,g+1 ,... ., P _M,g+1 and select a particle from it as the reconstruction plan to construct the g+1th microphone array.

From the above description, it can be found that the particle P _m,1 (1≤m≤M) continuously evolves and iterates through calculation, and generates particles P _m,2 , P _m,3 ,..., then in the gth construction After the microphone array ends and before the g+1th microphone array construction begins, the particle P corresponding to max(T _{m, 1} , T _{m, 2} ,..., T _{m, g} ) is the historical maximum. Optimal coordinate P _{m, best} , at this time max (T _{1, 1} , T _{1, 2} ,..., T _1,g , T 2, 1, T ₂ , ₂ ,..., The particle P corresponding to T _2,g ,...,TM _,1 , _TM,2 ,..., _TM,g ) is the global historical optimal coordinate G _best . It can be found that as the value of g increases, the value of P _{m, best} may change or remain unchanged, but the total number of P _{m, best} is always M.

in,

Among them, κ ₁ and κ ₂ are random numbers greater than 0 and less than 1, and α is the expansion coefficient. The historical optimal coordinates P _{m, best} and the global historical optimal coordinates G _best are used to iteratively update the particles to ensure that as the microphone array As the number of reconstructions increases, the entire particle evolves in the direction of gradually increasing signal gain ratio. On this basis, find the maximum value T f, g in T _{1, g} , T _{2, g} ,..., T _M _{, g} , M≥f≥1, and its corresponding particle is P _{f, g} , using P _{f, g+1} as the construction plan for the g+1th microphone array can minimize the impact of the multipath effect on the reconstructed microphone array.

Generally speaking, α is a random number, but in this embodiment α is a function. Specifically, the maximum value T f,g in T _1,g ,T _2,g ,...,T _M, _g , then P _f,g is the optimal particle, and P _m,g is the current particle. , it can be seen from the above that the particle can be represented by the coordinates of its internal microphone, so the distance between P _{f, g} and P _{m, g}

If γ<1, then the expansion coefficient α=0.5+γ, if γ=1, then α=1.8. By controlling the value of α, we avoid generating P _1,g +1,P _2,g _{+1,... through P 1,g,P 2,g} , _...,PM,g _. , when P _M,g+1 is too large, the randomness is too large, and then the g+1th microphone array construction plan is generated and the microphone array is reconstructed. Compared with the gth construction, the microphone array can obtain greater The signal gain ratio also avoids α being too small, which results in too many microphone array reconstructions, so that the sound source direction can be obtained as quickly as possible. In the case of γ < 1, adjusting α (essentially the iteration step size) according to the particle distance γ to obtain the reconstruction plan can maximize the increase in the signal gain ratio before and after the microphone array is reconstructed, and ensure After the microphone array is reconstructed, the signal gain ratio will not decrease, that is, the particles will diverge.

Example 2:

In order to ensure that P _1,1 , P _2,1 ,..., P _M,1 do not have too much influence on each other when iteratively updating, it is necessary to limit the position change of the peripheral microphone in the particle. The position movement distance of the jth peripheral microphone before and after the microphone array reconstruction is Δd _j , Δd _j <d _min .

Example 3:

P _1,g ,P _2,g ,...,P _M,g are called particles of the gth generation. Particles with the same g are called particles of the same generation, and the characteristic frequencies of particles of the same generation are the same. Which specific particle to select as the reconstruction solution for the microphone array from particles of the same generation is judged by the signal gain ratio. In Embodiment 1, among particles of the same generation, the particle with the largest signal gain ratio in the propagation direction corresponding to a single characteristic frequency sound is used as the reconstruction plan. In contrast, in this embodiment, for particles of the same generation, the signal gain ratio of a single particle in the corresponding propagation direction of each characteristic frequency sound is solved and then the average value is obtained, and then the average signal gain ratio of a single particle is obtained, and then The particle with the largest signal gain ratio to the average value is used as the reconstruction plan.

It is worth noting that if the average signal gain ratio is used as the selection criterion for the reconstruction scheme, the value of q should not be too large, otherwise it will cause the particles to diverge, that is, there is no guarantee that the signal gain ratio can increase after the particles are iteratively updated, so q is generally In terms of 2 or 3 or 4 or 5.

In particular, the average value of the signal gain ratio in this embodiment is a weighted average. The smaller the value of the characteristic frequency, the smaller the weight of the signal gain ratio of the sound signal in the corresponding propagation direction when calculating the weighted average. The reason is that the multipath effect has a relatively small impact on the low-frequency part of the sound signal, while the high-frequency part has a significant impact. Therefore, adding weight to the signal gain ratio corresponding to the higher characteristic frequency, and then realizing particle iteration, can make the average signal gain ratio Value increases more quickly.

The above are only specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. Those skilled in the art should understand that the present invention includes, but is not limited to, the drawings and those described in the above specific embodiments. content. Any modifications that do not depart from the functional and structural principles of the invention are intended to be included in the scope of the claims.

Claims

A sound source localization method under strong multipath interference conditions, which is characterized by including the following steps:

Step S1: Build a microphone array;

Step S2: The microphone array collects the sound source signal to determine the characteristic frequency ω p of the sound source signal, p = 1, 2,..., q, and filter the sound source signal to retain the ω p neighborhood range. The sound signal within ω p is then obtained, and the sound propagation direction corresponding to ω p is obtained;

Step S3: Calculate the signal gain ratio of the microphone array in the sound propagation direction corresponding to ω p ;

Step S4: Select the reconstruction scheme of the microphone array through the signal gain ratio, repeat steps S1 and S2 according to the reconstruction scheme to change ω p , and set the threshold δ. If the change of ω p is less than δ, then according to the changed The sound propagation direction corresponding to ω p is used to determine the sound source position. If the change in ω p is not less than δ, continue to repeat steps S3, S1 and S2 in sequence until the change in ω p is less than δ.
The sound source localization method under strong multipath interference conditions according to claim 1, characterized in that q is 2 or 3 or 4 or 5, and the value of q remains unchanged before and after steps S1 and S2 are repeated.
The sound source localization method under strong multipath interference conditions according to claim 1, characterized in that the microphone array includes N microphones, one of which is a central microphone, and the other microphones are peripheral microphones in sequence along the circumferential direction of the central microphone. Arrange, use the center microphone as the origin to establish spherical coordinates. The elevation angle of any point in the spherical coordinates connected to the center microphone in the spherical coordinates is
And the horizontal angle is θ, the distance between this point and the j-th peripheral microphone is r j , and the elevation angle of the line connecting this point and the j-th peripheral microphone in spherical coordinates is
And the horizontal angle is θ j , set the function
in
v is the speed of sound, i is the imaginary unit, and the elevation angle of the sound propagation direction corresponding to ω p in spherical coordinates is
And the horizontal angle is θ p , then the signal gain ratio in the sound propagation direction corresponding to ω p
The sound source localization method under strong multipath interference conditions according to claim 3, characterized in that when the microphone array is first constructed, the peripheral microphones are evenly distributed along the circumferential direction of the central microphone. At this time, the distance between two adjacent peripheral microphones is d. min , the position movement distance of the j-th peripheral microphone before and after the microphone array reconstruction is Δd j , Δd j <d min .
The sound source localization method under strong multipath interference conditions according to claim 3, characterized in that the multi-objective optimization algorithm of quantum particle swarm is used to generate the reconstruction plan of the microphone array to ensure that the microphone array is reconstructed after each reconstruction. The change in p decreases.
The sound source localization method under strong multipath interference conditions according to claim 5, characterized in that in step S3, the signal gain ratio is calculated for the sounds of multiple characteristic frequencies in their respective propagation directions and the weighted average is calculated, and the step S4 The reconstruction scheme is selected by a weighted average of signal gain ratios.
The sound source localization method under strong multipath interference conditions according to claim 6, characterized in that each microphone array arrangement is regarded as a particle, and the distance between the current particle and the optimal particle is γ, if γ< 1, then the expansion coefficient α=0.5+γ, if γ=1, then α=1.8.