阅读说明：本技术 一种采用听音定位技术实时监测目标物体运动轨迹的方法 (Method for monitoring motion trail of target object in real time by adopting listening positioning technology ) 是由 陈文兵 车文刚 于 2020-07-14 设计创作，主要内容包括：一种采用听音定位技术实时监测目标物体运动轨迹的方法,涉及军事、无人驾驶、智能机器人等技术领域。包括如下步骤：(1)用空间几何数学模型、传播过程中音强与声音传播距离的平方成反比、以及能量守恒定理等三种不同的方法建立微分方程来求解声源点(p)到各支麦克风(M<Sub>i</Sub>)的距离d<Sub>i</Sub>；(2)设计麦克风数量少且运算量小的克风分布树Tr；(3)采用最小二乘法从误差解集合中确定最优声源点位置p<Sub>i</Sub>；(4)由惠更斯原理采用回声定位方法确定各支麦克风的相对位置；(5)随机采样已解声源点的位置来拟合目标物体的运动曲线。(A method for monitoring a motion trail of a target object in real time by adopting a listening positioning technology relates to the technical fields of military affairs, unmanned driving, intelligent robots and the like. The method comprises the following steps: (1) three different methods, namely a space geometric mathematical model, the inverse proportion of the sound intensity to the square of the sound propagation distance in the propagation process, the energy conservation theorem and the like are used for establishing a differential equation to solve the problem from a sound source point (p) to each microphone (M) i ) Distance d of i (ii) a (2) Designing a small number of microphones and a small amount of calculation for a microphone distribution tree Tr; (3) determining the optimal sound source point position p from the error solution set by adopting a least square method i (ii) a (4) Determining the relative position of each microphone by adopting an echo positioning method according to a Huygens principle; (5) the positions of the de-noised source points are randomly sampled to fit the motion curve of the target object.)

1. A method for monitoring the motion trail of a target object in real time by adopting a listening positioning technology is characterized by comprising the following steps:

(1) solving the sound source point p to each microphone M_iDistance d of_i

The coordinates M of a plurality of points are known in a space rectangular coordinate system_i(x_i，y_i，z_i) Finding an unknown point p (x, y,0) and using the distance d from the known point M to the point p_iEstablishing an equation set to solve two unknown quantities x and y, wherein an unknown point p is an intersection point of two equation curves;

(2) designing a small number of microphones and a small number of calculations of a microphone distribution tree Tr

If the error analysis and optimization are not considered, only M is needed when the solving complexity ratio of the relative position of the microphone is comprehensively considered under the condition that the known object is on the horizontal ground₁，M₂，M₃The problem can be solved by three microphones, and when the three microphones are vertical to each other, the cost is saved most and the use requirement can be met;

(3) determining the optimal sound source point position p from the error solution set by adopting a least square method_i

In practical situations, sound is affected by noise and reverberant sound during propagation, and delay error exists when the sound propagates to two different microphones, so that

in practical situations, when the number of microphones is larger (greater than 3), the error space is smaller until an optimal solution appears, but the cost efficiency problem is considered, the microphones are appropriately added to construct a solution space, then the error processing is converted into a point problem for solving the optimal solution in a solution set H, the distance between three curves and the minimum point are used as the optimal solution point position through priori knowledge, W is a matrix for solving parameters of a linear equation set, X is a dependent variable matrix, and H is an independent variable, then the linear equation set of the ideal solution can be simplified as follows:

WX+b＝h (2)

the actual error solution equation set for point p can be expressed as:

e＝h-(WX+b) (3)

the solution of equation (3) will form a solution region due to the influence of external factors and measurement errors; therefore, a new method for reducing errors is to find a point (optimal point) with the minimum distance from one to three curves from a solution collection point space H, and researches find that the solution collection point space is necessarily in an area where the three curves simultaneously protrude outwards, and the optimal point is necessarily in a triangle formed by directly connecting three intersection points;

the coordinates of three vertexes of the triangle can be solved by solving the equation set, the linear equation of three sides of the triangle can be obtained by conversion, and the integrated midpoint p (x) is solved_i，y_i) In the following manner, when X is from X₁To X₃Taking x as each K cm interval jump_iWhile y is at f₁(x_i) And f₂(x_i) Or f₁(x_i) And f₃(x_i) The value is taken at intervals of K cm, if the interval is smaller, the error is lower, but the calculated amount is increased, and all error points are traversed in the mode;

from the above model, p (x)_i，y_i) (i-1, 2, …, n) using_i，kThe distance from the ith point to the kth curve is represented and can be solved by a point-to-straight line distance formula;

(4) determining relative position of each microphone by Huygens principle and echo positioning method

When the coordinates of the microphone relative to the hall are solved, an echo positioning model is established by using a sub sound source emitted by the microphone; in a closed space, a sub sound source receives two types of reflected waves, namely a vertical reflected wave and an oblique reflected wave; based on the priori knowledge and the derivation of the solid model, the diffused sound wave reflection track from the microphone M can be concluded as follows;

A. the oblique angle reflected wave can not return to the position M, and can do reciprocating motion for many times in space even if the oblique angle reflected wave can return;

B. the two reflected waves colliding linearly and the two reflected waves colliding perpendicularly have different "reverberant sound" wave fronts;

C. the microphone firstly receives 6 vertical reflected waves vertical to the wall surface, and the position of the microphone relative to the hall can be determined by taking the path of the vertical reflected waves;

when the microphone M_kHaving received 6 groups 1/2k of frequency echoes (k being the microphone number), the round trip time is Δ T₁，ΔT₁，…，ΔT₆By flat spectrum analysis of the "reverberant tone" to yield Δ T₁And Δ T₃、ΔT₂And Δ T₄、ΔT₄And Δ T₆The round trip time difference of three groups of linear impact sounds, and delta T₁、ΔT₂And Δ T₄The round-trip time difference of the echoes returning to M earlier in each group of collisions, so the spatial angle formed by the vertical round-trip lines of the echoes is the closest wall angle to the microphone tree;

solving for microphone M₁，M₂，…，M_nThe position of (2) is determined by using a microphone to perform sound positioning back and forth to determine a wall corner as a coordinate origin, and other microphones M_iThe positions of the microphones are all positioned on the basis of a space coordinate system determined by the first microphone;

from the above formula, M can be determined₁(x₁，y₁，z₁) Coordinates and coordinate systems of; all microphones M_iAll on the same microphone stand, as long as all microphones are not on the same line, the solving of the sound source position p will not be affected at last, therefore, in order to improve the calculation efficiency, the microphone tree can be improvedMiddle M₁Must place a microphone M vertically above₂Any other microphone may have to be connected to the microphone M₁(or M)₂) On the same horizontal plane; m₂The coordinates of (c) may be determined as:D_i，jis the distance between the two microphones;

M_ihas a vertical coordinate of z₁Thus finding the microphone M_i(x_i，y_i，z₂) Coordinates, requiring only the solution x_i，y_iTwo unknowns;

from equation set (5), x can be solved_i，y_iThereby determining M_i(x_i，y_i，z₂) The coordinates of (a); and z is₁And z₂That is the height of the microphone relative to the horizontal ground; the relationship between the coordinates of the sound source p and the coordinates of the microphone M is obtained previously, and the microphone M is connected to the sound source p_i(x_i，y_i，z₂) The backward substitution can determine the position of the target object in the hall;

(5) randomly sampling the positions of the de-noised source points to fit a motion curve of the target object so as to predict and reduce the calculation amount;

the random time points are taken to solve the position of the target object once, and the m sample points p are solved through the steps as the target object moves on the horizontal plane_i(x_i，y_i) (i ═ 1,2, …, m), solving a motion track curve function f (x) of the object, and converting the average speed v of the object into a function f (t) related to time, so that the initial time and the initial point position of the first received sound by the microphone can be used for predicting the next driving position of the object by the time t;

as can be seen from taylor's equation, any n-th order derivative function can be developed at x ═ 0:

equation (6) can be simplified as:

f(x)＝c₀+c₁x+c₂x²+c₃x³+…+c_nxⁿ(7)

solving for C (C) from equation (7)₀，c₁，c₂，…，c_n) F (x) can be solved, and when the number of sampling points m is more than n, f (x) can be obtained by solving an equation set; determination of C (C)₀，c₁，c₂，…，c_n) Using least squares criterion, m points p_i(x_i，y_i) Distance from curve f (x)_iThe sum of squares of (a) is minimal;

recording:

the inertia of an object exists in the motion process, the object generally travels in a straight line (at this time, linear regression simulation can also be used) f (x) is a linear function, the motion track of the object is most complicated to be a unitary cubic function at this time, and therefore m points are substituted into an equation set formed by (7), a solution of a group C with m-n (n is less than or equal to 3) is shared, the parameters of the group m-n are respectively substituted into f (x), and the parameters of the group m-n are substituted into a formula (8) to find an optimal group C (C)₀，c₁，c₂，c₃)；

The quadratic equation of one is: f (x) ═ c₀+c₁x+c₂x²，

The first order equation is: f (x) ═ c₀+c₁x+c₂x²+c₃x³。

2. The method of claim 1 wherein the target object is monitored in real time using a listening location techniqueThe motion trail method is characterized in that: solving for d by using a spatial geometric mathematical model_i，1：

Let n microphones M₁，M₂，…，M_nThe time when the same sound source signal is received is t_iThen M is_iAnd M₁Time difference of receiving signals sent by the same sound source:

t_i，1＝|t_i-t₁| (9)

the speed of sound propagation in air is a constant value c, and the sound sources p are respectively connected to the microphones M₁And a microphone M_iHas a distance difference of d₁、d_i：

d_i，1＝c.t_i，1(10)

d_i，1＝d_i-d₁(11)

From equations (10) and (11), it can be derived:

in conjunction with equation (1) can be derived:

the formula (13) is obtained by subtracting the following formula from the above formula:

wherein

x, y, d in the system of linear equations (15)₁For unknown quantity, combining the formula (1), establishing two relational expressions to solve x and y; when the number of microphones is 3, two TDOA measurements can be obtained, assuming first that d is₁Knowing, then, the location (x, y,0) of the sound source can be expressed in the form of a matrix multiplication of a system of linear equations:

the two sides are simultaneously multiplied by the inverse of the first matrix in equation (16) to obtain

3. A method for real-time monitoring of a target object's motion trajectory using a listening location technique, as claimed in claim 1, wherein: solving d by inverse proportion of sound intensity in propagation process to square of sound propagation distance_i：

When sound propagates in a constant medium, the sound intensity V at the microphone and the distance d to the sound source squared²In inverse proportion, there are:

V＝V₀/d²(17)

thus, the sound intensity V of the sound source at the two microphones is measured₁，V_iD can be solved by simultaneous equations₁，d_iIn the generation of equation (1), the coordinates p (x, y,0) of the sound source point can be solved.

4. A method for real-time monitoring of a target object's motion trajectory using a listening location technique, as claimed in claim 1, wherein: solving for d by the theorem of conservation of energy_i：

The sound follows the energy conservation theorem in the transmission process, theta is a factor set of energy reduction caused by the transmission of the sound in the air, deltad represents the linear transmission displacement of the sound, and a differential equation is established:

Δw＝-θ.Δd (19)

the simultaneous integration on both sides of differential equation (19) is:

∫dw＝∫-θdx (20)

solving (20) we can get the curve of sound as a function of displacement during propagation:

w＝w₀-θx (21)

where w0 is the initial energy of the sound source and x is the linear displacement of the acoustic signal in the current direction; since the sound emission time and w0 are unknown, only the energy w transmitted to the microphone by the same sound source can be measured_iNow assume that the energy w of the sound of three microphones is measured₁，w_i；

Matrix operation mode in computer:

wherein the third formula of equation (22) can be derived from equation (9) by solving the system of equations to find x₁，x_i(d_i) Substituting into equation (1) can find the sound source point coordinates p (x, y, 0).

Technical Field

The invention relates to the technical field of target object motion trail monitoring, in particular to a method for monitoring a target object motion trail in real time by adopting a listening positioning technology.

Background

The listening position model has important application in military and commercial fields. The listening location model is hopefully used in the direction of artificial intelligence, because the image is influenced by various uncertain factors in the shooting process, the processing and analysis technology of the image is difficult, the data volume is large, and even the recent deep learning method for comparing the heat is limited by the characteristics of data diversity, large calculation amount and the like. Therefore, listening to the localization model can simplify some problems that are difficult to analyze in three-dimensional space, such as aspects of automatic driving technology, robot work, etc. Meanwhile, the technology is combined with medical technology to improve the life style of the disabled.

Compared with the existing echo positioning method, the listening positioning technology does not need to calculate the position of a target object by actively sending sound by a detection object and then receiving the echo, the principle of the listening positioning technology is the same as that of an animal which utilizes ears to distinguish the direction of a sound source, and the direction of the sound source is judged by the sound received by two or more microphones at different positions. The method mainly establishes a differential equation analysis problem from different theoretical directions such as geometric mathematics, kinematic formulas, energy conservation theorem and the like, and improves an error optimization mode of the DTOA algorithm by using a least square method. And finally, fitting the motion trail of the target object by the sound sampled at the random time point of the microphone tree, so that the resources are saved, and the calculation amount is reduced.

Disclosure of Invention

The invention aims to provide a method for monitoring the motion trail of a target object in real time by adopting a listening positioning technology, which can realize the real-time monitoring of the motion state of the object in any closed space.

In order to achieve the purpose, the invention adopts the following technical scheme:

a method for monitoring the motion trail of a target object in real time by adopting a listening positioning technology comprises the following steps: (1) three different methods are adopted to establish a differential equation to solve the sound source point p to each microphone M_iDistance d of_i1), a space geometric mathematical model, 2), the sound intensity is inversely proportional to the square of the sound propagation distance in the propagation process, and 3) the energy conservation theorem; (2) designing a small number of microphones and a small amount of calculation for a microphone distribution tree Tr; (3) determining the optimal sound source point position p from the error solution set by adopting a least square method_i(ii) a (4) Determining the relative position of each microphone by adopting an echo positioning method according to a Huygens principle; (5) the positions of the de-noised source points are randomly sampled to fit the motion curve of the target object so as to make prediction and reduce the calculation amount.

The method further comprises the following steps:

(1) solving the sound source point p to each microphone M_iDistance d of_i

The coordinates M of a plurality of points are known in a space rectangular coordinate system_i(x_i,y_i,z_i) Finding an unknown point p (x, y,0) and using the distance d from the known point M to the point p_iEstablishing an equation set to solve two unknown quantities x and y, wherein an unknown point p is an intersection point of two equation curves;

1) solving for d by using a spatial geometric mathematical model_i,1：

Let n microphones M₁,M₂,…,M_nThe time when the same sound source signal is received is t_iThen M is_iAnd M₁Time difference of receiving signals sent by the same sound source:

t_i,1＝|t_i-t₁| (9)

the speed of sound propagation in air is a constant value c, and the sound sources p are respectively connected to the microphones M₁And a microphone M_iHas a distance difference of d₁、d_i：

d_i,1＝c.t_i,1(10)

d_i,1＝d_i-d₁(11)

From equations (10) and (11), it can be derived:

in conjunction with equation (1) can be derived:

the formula (13) is obtained by subtracting the following formula from the above formula:

wherein1,2, …, n; let x_i,1＝x_i-x₁,y_i,1＝y_i-y₁The target is located on the ground plane, so z is 0, and the reduction (14) has:

x, y, d in the system of linear equations (15)₁For unknown quantity, combining the formula (1), establishing two relational expressions to solve x and y; when the number of microphones is 3, two TDOA measurements can be obtained, assuming first that d is₁Knowing, then, the location (x, y,0) of the sound source can be expressed in the form of a matrix multiplication of a system of linear equations:

the two sides are simultaneously multiplied by the inverse of the first matrix in equation (16) to obtain

Substituting it into formula (1)

In (1), a value of d can be obtained₁The position of the sound source p can be determined by substituting the positive root of the quadratic equation (16).

2) Solving d by inverse proportion of sound intensity in propagation process to square of sound propagation distance_i：

When sound propagates in a constant medium, the sound intensity V at the microphone and the distance d to the sound source squared²In inverse proportion, there are:

V＝V₀/d²(17)

thus, the sound intensity V of the sound source at the two microphones is measured₁，V_iD can be solved by simultaneous equations₁，d_iIn the generation of equation (1), the coordinates p (x, y,0) of the sound source point can be solved.

3) Solving for d by the theorem of conservation of energy_i：

The sound follows the energy conservation theorem in the transmission process, theta is a factor set of energy reduction caused by the transmission of the sound in the air, deltad represents the linear transmission displacement of the sound, and a differential equation is established:

Δw＝-θ.Δd (19)

the simultaneous integration on both sides of differential equation (19) is:

∫dw＝∫-θdx (20)

solving (20) we can get the curve of sound as a function of displacement during propagation:

w＝w₀-θx (21)

where w0 is the initial energy of the sound source and x is the linear displacement of the acoustic signal in the current direction; since the sound emission time and w0 are unknown, only the energy w transmitted to the microphone by the same sound source can be measured_iNow assume that the energy w of the sound of three microphones is measured₁，w_i；

Matrix operation mode in computer:

wherein the third formula of equation (22) can be derived from equation (9) by solving the system of equations to find x₁,x_i(d_i) Substituting into equation (1) can find the sound source point coordinates p (x, y, 0).

(2) Designing a small number of microphones and a small number of calculations of a microphone distribution tree Tr

If the error analysis and optimization are not considered, only M is needed when the solving complexity ratio of the relative position of the microphone is comprehensively considered under the condition that the known object is on the horizontal ground₁，M₂，M₃The problem can be solved by three microphones, and when the three microphones are perpendicular to each other, the cost is saved most and the use requirement can be met.

(3) Determining the optimal sound source point position p from the error solution set by adopting a least square method_i

In practical situations, sound is affected by noise and reverberant sound during propagation, and delay error exists when the sound propagates to two different microphones, so that The model is put on a plane for analysis, wherein all curves meet at a point p in an ideal case, but all curves have an error intersection space in a large probability case, wherein all points form a solution set H;

in practical situations, when the number of microphones is larger (>3), the error space is smaller until an optimal solution occurs, but considering the cost efficiency problem, the microphones are appropriately added to construct a solution space, then the error processing is converted into a point problem for solving the optimal solution in a solution set H, the distance between three curves and the minimum point are used as the optimal solution point position through priori knowledge, W is a matrix for solving parameters of a linear equation set, X is a dependent variable matrix, and H is an independent variable, then the linear equation set of an ideal solution can be simplified as follows:

WX+b＝h (2)

the actual error solution equation set for point p can be expressed as:

e＝h-(WX+b) (3)

the solution of equation (3) will form a solution region due to the influence of external factors and measurement errors; therefore, a new method for reducing errors is to find a point (optimal point) with the minimum distance from one to three curves from a solution collection point space H, and researches find that the solution collection point space is necessarily in an area where the three curves simultaneously protrude outwards, and the optimal point is necessarily in a triangle formed by directly connecting three intersection points;

by calculatingSolving the equation set can solve the coordinates of three vertexes of the triangle, converting to obtain the linear equation of three sides of the triangle, and solving the integrated midpoint p (x)_i,y_i) In the following manner, when X is from X₁To X₃Taking x as each K cm interval jump_iWhile y is at f₁(x_i) And f₂(x_i) Or f₁(x_i) And f₃(x_i) The value is taken at intervals of K cm, if the interval is smaller, the error is lower, but the calculated amount is increased, and all error points are traversed in the mode;

from the above model, p (x)_i,y_i) (i-1, 2, …, n) using_i,kThe distance from the ith point to the kth curve is represented and can be solved by a point-to-straight line distance formula.

(4) Determining relative position of each microphone by Huygens principle and echo positioning method

When the coordinates of the microphone relative to the hall are solved, an echo positioning model is established by using a sub sound source emitted by the microphone; in a closed space, a sub sound source receives two types of reflected waves, namely a vertical reflected wave and an oblique reflected wave; based on the priori knowledge and the derivation of the solid model, the diffused sound wave reflection track from the microphone M can be concluded as follows;

A. the oblique angle reflected wave can not return to the position M, and can do reciprocating motion for many times in space even if the oblique angle reflected wave can return;

B. the two reflected waves colliding linearly and the two reflected waves colliding perpendicularly have different "reverberant sound" wave fronts;

C. the microphone firstly receives 6 vertical reflected waves vertical to the wall surface, and the position of the microphone relative to the hall can be determined by taking the path of the vertical reflected waves;

when the microphone M_kHaving received 6 groups 1/2k of frequency echoes (k being the microphone number), the round trip time is Δ T₁,ΔT₁,…,ΔT₆By flat spectrum analysis of the "reverberant tone" to yield Δ T₁And Δ T₃、ΔT₂And Δ T₄、ΔT₄And Δ T₆The round trip time of three groups of linear collision soundDifference, and Δ T₁、ΔT₂And Δ T₄The round-trip time difference of the echoes returning to M earlier in each group of collisions, so the spatial angle formed by the vertical round-trip lines of the echoes is the closest wall angle to the microphone tree;

solving for microphone M₁,M₂,…,M_nThe position of (2) is determined by using a microphone to perform sound positioning back and forth to determine a wall corner as a coordinate origin, and other microphones M_iThe positions of the microphones are all positioned on the basis of a space coordinate system determined by the first microphone;

from the above formula, M can be determined₁(x₁,y₁,z₁) Coordinates and coordinate systems of; all microphones M_iAll on the same microphone stand, as long as all microphones are not on the same line, the solving for the sound source position p will not be affected at last, therefore, in order to increase the computational efficiency, the microphone tree can be improved, where M is₁Must place a microphone M vertically above₂Any other microphone may have to be connected to the microphone M₁(or M)₂) On the same horizontal plane; m₂The coordinates of (c) may be determined as:D_i,jis the distance between the two microphones;

M_ihas a vertical coordinate of z₁Thus finding the microphone M_i(x_i,y_i,z₂) Coordinates, requiring only the solution x_i,y_iTwo unknowns;

from equation set (5), x can be solved_i,y_iThereby determining M_i(x_i,y_i,z₂) The coordinates of (a); and z is₁And z₂Is the opposite of the microphoneHeight above horizontal ground; the relationship between the coordinates of the sound source p and the coordinates of the microphone M is obtained previously, and the microphone M is connected to the sound source p_i(x_i,y_i,z₂) The back substitution may determine the location of the target object in the lobby.

(5) Randomly sampling the positions of the de-noised source points to fit a motion curve of the target object so as to predict and reduce the calculation amount;

the random time points are taken to solve the position of the target object once, and the m sample points p are solved through the steps as the target object moves on the horizontal plane_i(x_i,y_i) (i ═ 1,2, …, m), solving a motion track curve function f (x) of the object, and converting the average speed v of the object into a function f (t) related to time, so that the initial time and the initial point position of the first received sound by the microphone can be used for predicting the next driving position of the object by the time t;

as can be seen from taylor's equation, any n-th order derivative function can be developed at x ═ 0:

equation (6) can be simplified as:

f(x)＝c₀+c₁x+c₂x²+c₃x³+…+c_nxⁿ(7)

solving for C (C) from equation (7)₀,c₁,c₂,…,c_n) F (x) can be solved, when the number of sampling points m>When n, solving the equation set to obtain f (x); determination of C (C)₀,c₁,c₂,…,c_n) Using least squares criterion, m points p_i(x_i,y_i) Distance from curve f (x)_iThe sum of squares of (a) is minimal;

recording:

the inertia of an object exists in the motion process, the object generally travels in a straight line (at this time, linear regression simulation can also be used) f (x) is a linear function, the motion track of the object is most complicated to be a unitary cubic function at this time, and therefore m points are substituted into an equation set formed by (7), a solution of a group C with m-n (n is less than or equal to 3) is shared, the parameters of the group m-n are respectively substituted into f (x), and the parameters of the group m-n are substituted into a formula (8) to find an optimal group C (C)₀,c₁,c₂,c₃)；

The quadratic equation of one is: f (x) ═ c₀+c₁x+c₂x²，

The first order equation is: f (x) ═ c₀+c₁x+c₂x²+c₃x³。

Compared with the prior art, the invention has the following advantages: three different methods are used for determining the position relation of the target object relative to the microphone, and the microphone tree which is reasonably designed reduces the complexity of solving the coordinates of the microphone. The time delay error solution set caused by reverberation and noise is optimized by using a weighted least square method, and the error is reduced to a certain degree. And finally, predicting the motion trail of the human by curve fitting to reduce the calculated amount of the whole model.

Drawings

FIG. 1 is a flow chart of the present invention;

FIG. 2 is an optimal design of a microphone tree;

FIG. 3 is a schematic illustration of an idealized solution scenario;

fig. 4 is an error solution space set H.

Detailed Description

The present invention will be described in further detail with reference to the accompanying drawings.