CN104794894B - A kind of vehicle whistle noise monitoring arrangement, system and method - Google Patents

Abstract

The invention discloses a kind of vehicle whistle noise monitoring system and its method, the noise monitoring system includes blow a whistle Noise Identification module, noise source locating module, photographing module.Noise Identification of blowing a whistle module is less than or greater than the sound wave for noise of blowing a whistle, then detected to determine noise signal of blowing a whistle according to the frequency characteristic of vehicle whistle noise by adding band-pass filter to fall frequency in sound pick-up front end.More auditory localizations of the noise source locating module based on ESPRIT algorithms, the translation invariance for the signal subspace that ESPRIT algorithms are brought using the translation invariance of a kind of array, tried to achieve by constructing special pencil of matrix and generalized eigen decomposition being carried out to it with sound source to corresponding translation operator, to estimate so as to obtain directioin parameter.The noise source that photographing module pair determines seeks a shooting, and carries out violation mark processing, and then the violation vehicle photo of mark is stored or sends relevant traffic administrative department in real time.

Description

A car whistle noise monitoring device, system and method

technical field

The invention belongs to the technical fields of audio signal processing, pattern recognition and array signal processing, and relates to a car whistle monitoring device, system and method.

Background technique

Accurately and quickly judging the coordinate position of a moving sound source is the core goal of sound localization technology. At present, the microphone array sound source localization technology is the mainstream technology. It is a sound localization method that uses a microphone pick-up array to pick up voice signals and uses digital signal processing technology. It has good spatial selection characteristics. Based on this principle, there are three methods for sound source localization:

① Orientation technology based on high-resolution spectral estimation. It is a method of determining the direction angle of the sound source by solving the correlation matrix between the microphone signals, and further solving the position of the sound source. The method is derived from high-resolution spectral estimation techniques, such as minimum variance spectral estimation and eigenvalue decomposition. It requires the sound source signal to be stable, and in order to reduce the influence of external interference factors and meet the special conditions of the technology application, it is necessary to double the calculation amount of the system, thus putting forward higher requirements for the hardware equipment of the system .

② Steerable beamforming technology based on maximum output power. This technology processes the voice signal received by the microphone array to directly control the direction in which the microphone points to the maximum power beam of the sound source signal.

③ Sound source localization technology based on time difference of arrival (TDOA). The method is a wireless positioning technology, which realizes the positioning function by measuring the time difference of the sound source signal from the moving sound source arriving at each pickup. The ability to accurately and flexibly estimate the delay length is a key factor affecting sound source localization.

Contents of the invention

Based on the repeated bans of motor vehicle horns in densely populated urban areas and the low efficiency of manual surveillance. The objective of the invention is aimed at above-mentioned phenomenon, proposes a kind of automobile whistle noise monitoring system and method thereof. Detect illegal honking vehicles on roads where honking is prohibited, and use road cameras to find points and shoot.

The invention provides a car whistle noise monitoring system, which includes:

Whistle noise recognition module, which is used to identify the picked-up sound waves, and finally determine the car whistle noise signal;

Noise source location module, which determines the source of car whistle noise according to the car whistle noise signal identified by the whistle noise identification module;

The camera module is used for photographing the determined car horn noise source.

The present invention also provides a kind of car whistle noise monitoring method, it comprises:

Identify the picked up sound waves, and finally determine the noise signal of the car whistle;

Determine the car whistle noise source according to the car whistle noise signal identified by the whistle noise identification module;

Take pictures of the identified car horn noise sources.

The present invention also provides a kind of car whistle noise monitoring system, it comprises:

Whistle noise recognition module, which includes a plurality of pickup units, respectively used to pick up the sound waves of the surrounding environment, and identify the picked up sound waves, and finally determine the car whistle noise signal;

A noise source localization module, which includes a first microphone array and a second microphone array, the microphone array includes a first microphone array and a second microphone array respectively distributed on the X axis and the Y axis, and the first microphone array and the second microphone array The microphone array includes uniformly spaced equivalence microphone array elements; the first microphone array and the second microphone array determine the car whistle noise source according to the car whistle noise signal identified by the whistle noise identification module;

The camera module includes a plurality of cameras, which are respectively used to shoot the determined noise source of the car whistle.

The point-finding and shooting module of the road camera performs point-finding and shooting of the camera according to the three-dimensional points determined by the sound source localization module. Then mark the pictures taken, and save or transmit the pictures of the marked violating vehicles to the traffic management department in real time.

Description of drawings

Fig. 1 is the principle block diagram of automobile whistle noise monitoring system in the present invention;

Fig. 2 is the system diagram of automobile whistle recognition system among the present invention;

Fig. 3 is the distribution structure diagram of array array source in the present invention;

Fig. 4 installation diagram of the system device in the present invention.

detailed description

In order to make the object, technical solution and advantages of the present invention clearer, the present invention will be further described in detail below in conjunction with specific embodiments and with reference to the accompanying drawings.

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

Fig. 1 is the structural block diagram of the automobile whistle noise monitoring system that the present invention proposes. As shown in Figure 1, the system includes three modules, namely whistle noise identification module, noise source location module, and camera module. The car whistle noise monitoring system proposed by the present invention uses a microphone array to pick up sound source signals, and then processes the picked up sound source signals by a central processing unit.

The recognition module includes a plurality of sound pickup units, which are used to perform band-pass filtering on the picked-up sound waves, to filter out road noises whose frequencies are lower and higher than car whistles, and only keep audio signals near the whistles. Then, the recognition module performs short-term energy over-limit detection on the audio signal near the retained whistle sound, and detects a suspected car whistle signal. Finally, frequency matching is performed to determine the whistle signal.

Optionally, the plurality of sound pickup units perform band-pass filtering on the picked-up sound waves according to the upper and lower threshold values, and the upper and lower threshold values can be determined according to the statistics of the whistle frequencies of general motor vehicles. Optionally, according to the frequency characteristics of the motor vehicle whistle sound measured in the experiment, a band-pass filter with a cutoff frequency of 200Hz and 5KHz is added to the front end of the microphone sensor, which can filter out the road noise to the greatest extent while retaining the mid-frequency part of the car Whistle signal component.

Wherein, analog-to-digital conversion is performed on the processed band-pass signal to further determine whether it is whistle noise.

The identification module of the present invention also performs short-term energy over-limit detection on the audio signal near the whistle sound obtained after band-pass filtering. This is because, due to the time-varying characteristics of the signal, the window function is used to divide the signal into frames of a certain length, and it is considered that the characteristics of the signal are basically unchanged in these short periods of time. In the present invention, a relatively smooth frequency spectrum is obtained by using a Hamming window with a wider main lobe.

The invention adopts the endpoint detection method based on short-time energy in the time domain, and the purpose of the endpoint detection is to detect the position of the voice signal that is suspected to be the whistle of a car.

The short-term energy of a frame of sound signal x(n) is defined as:

Among them, w(n) is the window function, N is the window length, which is also the length of a sound processing frame, and x(n) and x(m) are the frame signals at time n and m. After the first step of band-pass filtering the sound signal, most of the road noise is removed. Here, the detection threshold can be set as E _l . When E _n > E _l , it is determined that there is a suspected car whistle in the current sound sub-frame signal appears.

Wherein, E _l =min[0.03(E _max -E _min )+E _r , 4E _r ]

Among them, E _max is the maximum value of energy, E _min is the minimum value of energy, and E _r is the average value of energy of all sub-frames.

After detecting the suspected car whistle signal frame, the identification module also performs frequency comparison to determine the whistle signal from the suspected car whistle signal frame. Specifically, the identification module transforms the suspected car whistle signal detected by the short-term energy method into the frequency domain, and uses the statistical characteristics of the whistle and noise spectrum to identify the whistle. That is, the pre-statistically determined whistle noise frequency characteristics are used to match it, and if matched, the suspected car whistle signal is considered to be a definite whistle signal.

In the present invention, the multi-noise source location module locates the sound source according to the whistle sound signal determined by the identification module. Optionally, the multi-noise source localization module utilizes the translation invariance of the signal subspace brought about by the translation invariance of a class of arrays based on the ESPRIT algorithm, and obtains the corresponding noise source by constructing a special matrix bundle and performing generalized eigendecomposition on it. The source comes from the corresponding translation operator, so as to obtain the direction parameter estimation. In this step, the three-dimensional coordinates of the signal source can be calculated based on the direction of the signal source determined for each sound pickup unit. The implementation process of this module will be introduced in detail below.

In the present invention, the multi-noise sound source localization module includes a first microphone array and a second microphone array installed inside the sound pickup device. Fig. 3 shows a schematic diagram of the array distribution of the first microphone array and the second microphone array in the present invention. As shown in Figure 3, the first microphone array and the second microphone array are distributed as uniform linear arrays X and Y with N array elements on the x-axis and y-axis respectively, and the array elements at the origin are two linear arrays X and Y are shared, the first microphone array and the second microphone array are composed of the same array elements, the first microphone array and the second microphone array are symmetrical about the plane, and the center positions of the two arrays are separated by L, and L is set according to actual needs .

Optionally, in the present invention, when the determined whistle signal is three-dimensionally positioned, the azimuth angle of the noise source relative to the array center point can be obtained based on the ESPRIT algorithm, assuming that the center of the first microphone array is the three-dimensional coordinate origin (0 , 0, 0), the array element spacing of the linear array is d, the output noise of the array is zero-mean, variance is σ ² statistically independent Gaussian white noise, and is not correlated with the noise source.

If there are M statistically independent noise sources in space, the frequencies and incident angles of the noise sources relative to the reference array element are respectively f _i , θ _i , For the signal vectors received by the linear array X and the linear array Y are respectively

Wherein: S (t)=[s _l (t), ..., s _k (t), ... s _M (t)] ^T is the whistle noise vector;

Represents the direction matrix of the line array X;

Represents the direction matrix of the line array Y;

a _x (f _i , θ _i , φ _i ) represents the receiving function for noise source i, and the direction matrix of different receiving array elements is different;

N _x (t) and N _y (t) respectively represent the noise signals other than whistle noise received by the line array.

In order to jointly estimate the two-dimensional angle of arrival and frequency of multiple whistle noise signals, the present invention is based on a joint estimation method of the ESPRIT algorithm, but the present invention does not limit that only this method can be applied to this system. In this microphone array, the received signal vectors of the first N-1 array elements and the last N-1 array elements of subarray X are denoted as X ₁ (t) and X ₂ (t) respectively, and the first N array elements of subarray Y The received signal vectors of -1 array element and the last N-1 array elements are respectively denoted as Y ₁ (t) and Y ₂ (t), then we have

in:

A _x is the direction matrix of the X array, and A _y is the direction matrix of the Y array.

To estimate the frequency of the signal, a delay τ is added to the received signal of the array.

It is only necessary to delay the first N-1 array elements in the X sub-array by one section, and the data obtained after the delay is recorded as X ₃ , which can be known according to the spatial two-dimensional spectrum estimation:

in

According to the principle of rotation invariant subspace, if the spatial noise is white noise, the autocovariance matrix of X ₁ (t) can be constructed Cross-covariance matrix of X ₁ (t) and X ₂ (t) Autocovariance matrix of Y ₁ (t) Cross-covariance matrix of Y ₁ (t) and Y ₂ (t) Cross-covariance matrix of X ₁ (t) and X ₃ (t) Assuming that the spatial noise is Gaussian white noise, and denoising the above five covariance matrices, we have

obtain separately Generalized eigenvalues for pairs of 3 matrices:

vf _i = exp(-j2πf _i τ),

Among them, i=1, 2, ..., M, M is the number of whistle noise sources.

Combining the above three equations, the frequency, azimuth and depression angle of the noise source can be solved:

vf _i ＝exp(-j2πf _i τ)

Among them, θi and _φi are the azimuth angle and overlooking angle of the _i -th noise source respectively; f _i is the frequency of the i-th noise source, and τ is the received signal delay of the receiving line array X ₃ relative to the receiving line array X ₁ ; d is the distance between array elements in the first microphone array and the second microphone array, c is the speed of light; vf _i , vX _i , vY _i are matrix pairs generalized eigenvalues of . In the process of solving the above equations, by constructing a special matrix bundle and performing generalized eigendecomposition on it, it is assumed that vf _i , vX _i , and vY _i are matrix The generalized eigenvalues of the matrix Expressed as follows:

in, is the autocovariance matrix of X ₁ (t), is the cross-covariance matrix of X ₁ (t) and X ₂ (t), is the autocovariance matrix of Y ₁ (t), is the cross-covariance matrix of Y ₁ (t) and Y ₂ (t), is the cross-covariance matrix of X ₁ (t) and X ₃ (t), I is the identity matrix, σ ² is the variance of Gaussian white noise; X ₁ (t) and X ₂ (t) are the first microphone array The received signal vectors of the first N-1 array elements and the last N-1 array elements of the sub-array X; Y ₁ (t) and Y ₂ (t) are respectively the first N-1 arrays of the sub-array Y of the first microphone array X ₃ (t) is the signal vector after the first N-1 elements in the sub-array X of the first microphone array delay the received signal by τ.

According to the received signal vectors of the corresponding elements of the first microphone array and the second microphone array, the spatial coordinates of the i-th noise source can be calculated

z _i =(z _i1 +z _i2 )/2

Among them, x _i , y _i and z _i are the spatial coordinates of the i-th noise source respectively; θ _i1 and are respectively the azimuth angle and pitch angle of the i-th noise source relative to the center point of the first microphone array; θ _i2 and are the azimuth angle and elevation angle of the i-th noise source relative to the center point of the second microphone array, respectively.

The camera module described in the present invention includes a plurality of cameras, which are evenly distributed between the first microphone array and the second microphone array of the multi-noise sound source localization module, as shown in Figure 4; when the multi-noise sound source localization module After obtaining the spatial coordinates of the whistle signal, the camera module controls the rotation and focus of the camera, and takes pictures of the noise source. Specifically, all cameras may be selected to be aimed at the noise source to shoot, or at least one camera with the closest distance may be selected to shoot at the noise source.

Among them, the photos taken by the camera are marked for violations.

Wherein, the marked photos are stored or transmitted to relevant traffic departments in real time.

Among them, the camera is adjusted to find and shoot the nearest point noise source.

Wherein, the number of cameras can be determined according to the honking frequency of cars in the road section.

Take pictures of cars that whistle in violation of regulations, and store the pictures in the camera memory, or transmit the pictures to relevant traffic management departments in real time.

Among them, the noise source photos taken are flagged for violation of the whistle.

Among them, the memory in the camera is regularly processed to prevent the memory from being too full and unable to shoot.

The specific embodiments described above have further described the purpose, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above descriptions are only specific embodiments of the present invention, and are not intended to limit the present invention. Within the spirit and principles of the present invention, any modifications, equivalent replacements, improvements, etc., shall be included in the protection scope of the present invention.

Claims (7)

1. A car whistle noise monitoring system, comprising: Whistle noise recognition module, which is used to identify the picked-up sound waves, and finally determine the car whistle noise signal; Noise source location module, which determines the source of car whistle noise according to the car whistle noise signal identified by the whistle noise identification module; The camera module includes a plurality of cameras, which are used to shoot the determined car whistle noise source, perform camera point-finding and shooting according to the three-dimensional points determined by the sound source localization module, and then identify the captured pictures, and place the identified illegal vehicle pictures Save or transmit to the traffic management department in real time; Wherein, the whistle noise identification module identifies the picked up sound waves as follows: Perform band-pass filtering on the picked-up sound waves to filter out other noise signals whose frequency is not outside the range of the car whistle noise signal; Perform short-term energy over-limit detection on the band-pass filtered sound wave, and detect the suspected car whistle noise signal; Matching the frequency of the suspected car whistle noise signal with the frequency of the preset car whistle noise signal to determine the car whistle noise signal; The short-term energy over-limit detection of the band-pass filtered sound wave specifically includes: Using a window function to divide the band-pass filtered sound wave into a plurality of frame signals, wherein the characteristics of the signal in each frame signal are unchanged; Calculate the short-term energy of each frame signal and compare it with the energy detection threshold; When the short-term energy of the frame signal is greater than the energy detection threshold, it is determined as a suspected car whistle noise signal; The short-term energy of each frame signal is calculated as follows: Among them, E _n is the short-term energy of the nth frame signal, w(n) is the window function, N is the window length, and x(n) is the nth frame signal; The energy monitoring threshold value is calculated as follows: E _l =min[0.03(E _max -E _min )+E _r ,4E _r ] Among them, E _max is the energy maximum value in all frame signals, E _min is the energy minimum value in all frame signals, E _r is the average value of all frame signal energy; Wherein, the noise source location module determines the car whistle noise source according to the car whistle noise signal identified by the whistle noise identification module: Acquire the signal vector of each car whistle noise source that the microphone array receives; Wherein, the microphone array includes a first microphone array and a second microphone array distributed on the X-axis and the Y-axis respectively, and the first microphone array and the second microphone array The two-microphone array includes equally spaced microphone array elements; According to the acquired signal of each car whistle noise source, the frequency, azimuth angle and overlooking angle of each car whistle noise source are estimated based on the ESPRIT algorithm; The spatial coordinates of each vehicle whistle noise source are calculated according to the frequency, azimuth angle and overlooking angle of each vehicle whistle noise source. 2. system as claimed in claim 1, wherein, described whistling noise identification module utilizes a plurality of pick-up devices to carry out the pick-up of sound wave; Described noise source localization module utilizes the microphone array that is installed in pick-up device inside to noise source Positioning; the camera module uses a plurality of cameras to shoot the noise source of the car whistle. 3. system as claimed in claim 2, wherein, the frequency of described each car whistle noise source, azimuth angle and the angle of view of looking down are calculated as follows: vf _i = exp(-j2πf _i τ), Wherein, θ _i and φ _i are the azimuth angle and the viewing angle of the i noise source respectively; f _i is the frequency of the i noise source, and τ is the time delay; d is the first microphone array and the second microphone array The distance between the array elements, c is the speed of light; vf _i , vX _i and vY _i are the calculation intermediate values respectively. 4. system as claimed in claim 1, wherein, the spatial coordinates of described each car whistle noise sources are calculated as follows: z _i =(z _i1 +z _i2 )/2 Among them, x _i , y _i and z _i are the spatial coordinates of the i-th noise source respectively; θ _i1 and are respectively the azimuth angle and pitch angle of the i-th noise source relative to the center point of the first microphone array; θ _i2 and are the azimuth and elevation angles of the i-th noise source relative to the center point of the second microphone array; L is the distance between the centers of the first microphone array and the second microphone array. 5. The system of claim 1, wherein the short-term energy of each frame signal is calculated as follows: Among them, E _n is the short-term energy of the nth frame signal, w(n) is the window function, N is the window length, and x(n) is the nth frame signal; The energy monitoring threshold value is calculated as follows: E _l =min[0.03(E _max -E _min )+E _r ,4E _r ] Among them, E _max is the energy maximum value in all frame signals, E _min is the energy minimum value in all frame signals, E _r is the average value of all frame signal energy. 6. A method for monitoring car whistle noise, comprising: Identify the picked up sound waves, and finally determine the noise signal of the car whistle; Determine the car whistle noise source according to the car whistle noise signal identified by the whistle noise identification module; Take pictures of the determined car horn noise source; Among them, the picked-up sound waves are identified as follows: Perform band-pass filtering on the picked-up sound waves to filter out other noise signals whose frequency is not outside the range of the car whistle noise signal; Perform short-term energy over-limit detection on the band-pass filtered sound wave, and detect the suspected car whistle noise signal; Matching the frequency of the suspected car whistle noise signal with the frequency of the preset car whistle noise signal to determine the car whistle noise signal; The short-term energy over-limit detection of the band-pass filtered sound wave specifically includes: Using a window function to divide the band-pass filtered sound wave into a plurality of frame signals, wherein the characteristics of the signal in each frame signal are unchanged; Calculate the short-term energy of each frame signal and compare it with the energy detection threshold; When the short-term energy of the frame signal is greater than the energy detection threshold, it is determined as a suspected car whistle noise signal; The short-term energy of each frame signal is calculated as follows: Among them, E _n is the short-term energy of the nth frame signal, w(n) is the window function, N is the window length, and x(n) is the nth frame signal; The energy monitoring threshold value is calculated as follows: E _l =min[0.03(E _max -E _min )+E _r ,4E _r ] Among them, E _max is the energy maximum value in all frame signals, E _min is the energy minimum value in all frame signals, E _r is the average value of all frame signal energy. 7. A car whistle noise monitoring system, comprising: Whistle noise recognition module, which includes a plurality of pickup units, respectively used to pick up the sound waves of the surrounding environment, and identify the picked up sound waves, and finally determine the car whistle noise signal; A noise source localization module, which includes a first microphone array and a second microphone array, the microphone array includes a first microphone array and a second microphone array respectively distributed on the X axis and the Y axis, and the first microphone array and the second microphone array The microphone array includes uniformly spaced equivalence microphone array elements; the first microphone array and the second microphone array determine the car whistle noise source according to the car whistle noise signal identified by the whistle noise identification module; A camera module, which includes a plurality of cameras, respectively used to shoot the determined car whistle noise source; Wherein, the whistle noise identification module identifies the picked up sound waves as follows: Perform band-pass filtering on the picked up sound waves to filter out other noise signals whose frequency is not outside the range of the car whistle noise signal; Perform short-term energy over-limit detection on the band-pass filtered sound wave, and detect the suspected car whistle noise signal; Matching the frequency of the suspected car whistle noise signal with the frequency of the preset car whistle noise signal to determine the car whistle noise signal; The short-term energy over-limit detection of the band-pass filtered sound wave specifically includes: Using a window function to divide the band-pass filtered sound wave into a plurality of frame signals, wherein the characteristics of the signal in each frame signal are unchanged; Calculate the short-term energy of each frame signal and compare it with the energy detection threshold; When the short-term energy of the frame signal is greater than the energy detection threshold, it is determined as a suspected car whistle noise signal; The short-term energy of each frame signal is calculated as follows: Among them, E _n is the short-term energy of the nth frame signal, w(n) is the window function, N is the window length, and x(n) is the nth frame signal; The energy monitoring threshold value is calculated as follows: E _l =min[0.03(E _max -E _min )+E _r ,4E _r ] Among them, E _max is the energy maximum value in all frame signals, E _min is the energy minimum value in all frame signals, E _r is the average value of all frame signal energy.