CN104991573A - Locating and tracking method and apparatus based on sound source array - Google Patents

Abstract

The present invention discloses a locating and tracking method based on a sound source array, comprising the following steps: S1, acquiring on-site sound through a quintuple microphone array, and pre-processing a sound signal acquired by each microphone in the quintuple microphone array to obtain an audio signal; S2, performing sound source locating for the audio signals according to arrival time delay of the audio signals between the microphones and positional information of the microphone array, so as to calculate a pitch angle, an azimuth angle and an object distance; and S3, moving and turning a locating and tracking apparatus to arrive to a sound source position. The locating and tracking method of the present invention performs related processing for influences of non-gauss noises, coherent noises and indoor reverberation of the sound source on accurate locating, thereby improving accuracy of locating of the sound source.

Description

A kind of positioning and tracing method based on acoustic source array and device thereof

Technical field

The present invention relates to a kind of positioning and tracing method based on acoustic source array, relate to a kind of positioning and tracking device based on acoustic source array simultaneously.

Background technology

Auditory localization technology utilizes microphone to pick up voice signal, and carry out treatment and analysis with Digital Signal Processing to it, then determines and follow the tracks of the locus of sound source.In the prior art, have high-performance, high directivity single microphone when noiseless, without reverberation, distance sound source very near, high-quality sound-source signal can be obtained.But the pickup scope of single microphone is very limited, if sound source is outside the choice direction of microphone, then can introduces a large amount of noises, cause the Quality Down of pickoff signals; And the signal that single microphone receives, is superposed by multi-acoustical and neighbourhood noise and forms, cannot realize the separation of each sound source.In actual applications, due to sound source may indoor among a small circle in motion, the in addition factor such as the multipath reflection of indoor other sound various and reverberation, the Signal-to-Noise that single microphone also can be caused to receive reduction, the Quality Down of pickoff signals.

In order to solve the limitation of single microphone, someone proposes the auditory localization based on microphone array.It is a large study hotspot of field of voice signal in recent years.In conjunction with the needs of practical application, as Remote Video Conference, warning system, car phone, hearing assistance system, intelligent robot system etc., there has been proposed and put by certain geometry with multiple microphone, composition microphone array carries out the method for speech processes.It has very strong spatial selectivity, need not move and just automatically can monitor sound-source signal, locates and follow the tracks of, and therefore in speech enhan-cement, auditory localization, echo cancellor etc., has great application space.Wherein, the voice signal utilizing microphone array to collect is looked to the parent estimating sound source position and receive people.

But in the environment that auditory localization microphone array multidigit is comparatively closed in indoor etc., except the impact of non-Gaussian noise and coherent noise, this reverberation in indoor of sound source also can be aimed at and be determined that position impacts.Therefore, how in real sound field environment, to locate three-dimensional point sound source, positional information is accurately provided, become the active demand of people.

In order to solve the problem, be in the Chinese patent application of 201210204877.3 at application number, disclosing a kind of sonic location system based on microphone array.LabVIEW is applied to the exploitation of auditory localization software by this system, completes the Acquire and process of sound signal easily, uses broad sense cross correlation function method to carry out time delay estimation, sets up system of equations realize location in conjunction with planar four-element cross battle array model.But the reverberation of this system for non-Gaussian noise, coherent noise and sound source originally in indoor is aimed at and is determined that position impacts, and does not carry out the process of being correlated with.In addition, azimuth angle error, angle of pitch error, target range error that time delay is estimated to cause are all relevant with the velocity of sound, array element distance, the angle of pitch and position angle, have had a strong impact on positioning precision.

Summary of the invention

For the deficiencies in the prior art, primary technical matters to be solved by this invention is to provide a kind of positioning and tracing method based on acoustic source array.

Another technical matters to be solved by this invention is to provide a kind of positioning and tracking device based on acoustic source array.

For achieving the above object, the present invention adopts following technical scheme:

Based on a positioning and tracing method for acoustic source array, comprise the steps:

S1, adopts the on-the-spot sound of five yuan of microphone array acquisition, carries out pre-service obtain sound signal to the voice signal of microphone collection each in five yuan of microphone arrays;

S2, carries out the location of sound source, calculates the angle of pitch, position angle and target range to sound signal according to the positional information of the sound signal time delay of arrival between each microphone and microphone array;

S3, to move positioning and tracking device according to the angle of pitch calculated, position angle and target range and turns to, and arrives sound source position.

Wherein more preferably, in step sl, five microphone composition cross formations in described five yuan of microphone arrays, for picking up same sound-source signal.

Wherein more preferably, in step sl, voice signal carries out pre-service and obtains sound signal, comprises the steps:

S11, carries out windowing framing by voice signal, obtains the signal segment of short-term stationarity; Wherein, framing is with specific window function w (n) and voice signal s ₁n () carries out convolutional calculation, thus form windowing signal s _w(n)=s ₁(n) * w (n);

S12, carries out breaking point detection to voice signal, calculates short-time energy E, if E is less than threshold value can think noise, wherein, n is sampling number, otherwise, turn to step S13;

S13, calculates short-time zero-crossing rate Z ₀if, Z ₀in thresholding, then represent sound signal band noise being detected, the windowing signal s of the sound signal of band noise _wn the short-time zero-crossing rate of () is Z ₀: $Z_0=\frac{1}{2}\left\{\underset{n=1}{\overset{N-1}{\Σ}}\right|\mathrm{sgn}\[s_w\left(n\right)\]-\mathrm{sgn}\[s_w(n-1)\]\left|\right\};$

S14, after the sound signal of band noise detects, utilizes spectrum-subtraction denoising, and feeding FIR bandpass filter is carried out filtering process and obtained sound signal.

Wherein more preferably, in step S14, utilize spectrum-subtraction denoising, comprise the steps:

S141, carry out Fourier transform to the sound signal of band noise, obtain: X (ω)=S (ω)+N (ω), rated output is composed: | X (ω) ²|=| S (ω) | ²+ | N (ω) | ²+ S (ω) N ^*(ω)+S ^*(ω) N (ω); Wherein, S ^*(ω) N (ω) and S (ω) N ^*(ω) 0, P is _s(ω), P _x(ω) and P _n(ω) represent the power spectrum of s (t), x (t) and n (t) respectively, s (t) is sound signal, and n (t) is noise signal, and x (t) to make an uproar sound signal, then P for band _x(ω)=P _s(ω)+P _n(ω);

S142, gets the frame data before sound signal being detected as noise signal, obtains noise power spectrum P _n(ω);

S143, according to formula: P _x(ω)=P _s(ω)+P _n(ω) pure sound spectrum is obtained;

S144, carries out to the pure sound spectrum obtained the amplitude spectrum that evolution obtains pure audio;

S145, makes Fourier transform to sound signal s (t), obtains its phase place, by described phase multiplication with the amplitude spectrum of pure audio, then carries out Fourier inversion and obtains composing the sound signal after subtracting.

Wherein more preferably, in step s 2, according to the positional information of the sound signal time delay of arrival between each microphone and microphone array to the location of carrying out pretreated sound signal and carry out target sound source, comprise the steps:

S21, by broad sense cross correlation function method, calculate five yuan of microphone array received to sound signal in, the relative time delay between every two-way audio signal;

S22, according to the relative time delay that step S21 obtains, utilizes the geometric relationship of array element and sound source position to obtain the position of sound source;

The position of described sound source comprises the position angle of sound source, the angle of pitch and target range.

Wherein more preferably, in the step s 21, adopt broad sense cross correlation function method calculate five yuan of microphone array received to sound signal in, in the relative time delay between every two-way audio signal, comprise the steps:

S211, asks Fourier transform to two-way audio signal, gets conjugation to the data after a wherein road Fourier transform;

S212, calculates the cross-power spectrum between two-way audio signal;

S213, is transformed into time domain by frequency-region signal, obtains broad sense cross correlation function, and wherein, broad sense cross correlation function is: $R_{12}\left(\mathrm{\τ}\right)={\∫}_0^{\mathrm{\π}}{\mathrm{\ψ}}_{12}\left(\mathrm{\ω}\right)X_1\left(\mathrm{\ω}\right)X_2^{*}\left(\mathrm{\ω}\right)e^{-\mathrm{j\ωt}}d\mathrm{\ω};$ X ₁(ω), X ₂(ω) be x respectively ₁(t), x ₂the Fourier transform of (t), cross-power spectrum, ψ ¹²for weight function;

S214, finds the peak of broad sense cross-correlation related function, is the relative time delay between sound signal;

S215, filters the misdata in the relative time delay obtained.

Wherein more preferably, in step S215, the misdata in the relative time delay obtained is filtered, comprises the steps:

S2151, carries out sound detection, detects that sound then enters next step, sound do not detected in the schedule time, then empty time delay value array;

S2152, utilizes the acoustic segment comprised in time delay value array to calculate time delay value.

Wherein more preferably, in step S2152, calculate time delay value and comprise the steps:

S21521, compares time delay number and delay threshold in time delay value array, when time delay number is less than delay threshold in time delay value array, the time delay value newly estimated is put into array end;

S21522, when in time delay value array, time delay number equals delay threshold, then by-1 the time delay reach of delay threshold after array, then the time delay value calculated by new estimation puts into array end;

S21523, again judges time delay number in time delay value array, when being less than delay threshold, then turns to step S21521, when equaling delay threshold, then turn to step S21524;

S21524, sorts to the time delay value in time delay array, removes 3 maximal values and 3 minimum value, asks the average of residual time delay threshold value-6 values, this average delay is carried out the orientation of sound source as time delay value.

Based on a positioning and tracking device for acoustic source array, for realizing above-mentioned positioning and tracing method, comprise moving cell, control module, audio unit and display unit;

Wherein, described audio unit in order to collection site sound, and carries out pre-service to the voice signal of microphone collection each in five yuan of microphone arrays;

Sound signal after the process that described control module audio reception unit sends, is processed by auditory localization algorithm, and the auditory localization information after process is delivered to display unit, meanwhile, the positioning control order after process is sent to moving cell;

Described moving cell receives the control command from control module, and the positioning and tracking device being used for controlling acoustic source array carries out rotating and travelling according to orientation;

Described display unit receives the state of auditory localization information and the positioning and tracking device obtained after treatment from control part unit, displays it.

Wherein more preferably, described moving cell sends control signal to motor driver by control module, and drive motor completes corresponding go to action, makes positioning and tracking device based on acoustic source array towards the orientation of main sound source;

Described moving cell comprises two driving wheels, is arranged at the left and right sides of electric drive respectively, completes advance or backward movement when two driving wheels synchronously move at the uniform velocity, and when giving two driving wheel friction speeds, positioning and tracking device is rotated.

Positioning and tracing method based on acoustic source array provided by the present invention, adopts the on-the-spot sound of five yuan of microphone array acquisition, carries out pre-service to the sound signal of microphone collection each in five yuan of microphone arrays.Carry out the location of sound signal according to the positional information of the sound signal time delay of arrival between each microphone and microphone array, calculate the angle of pitch, position angle and relative distance.Reverberation for non-Gaussian noise, coherent noise and sound source originally in indoor is aimed at and is determined the process that the impact that position is caused is correlated with to significantly improve the accuracy of auditory localization.

Accompanying drawing explanation

Fig. 1 is the schematic diagram of the positioning and tracking device based on acoustic source array provided by the present invention;

Fig. 2 is the process flow diagram of the positioning and tracing method based on acoustic source array provided by the present invention;

Fig. 3 is existing based in the positioning and tracing method of acoustic source array, the planar structure schematic diagram of planar four-element array;

Fig. 4 is existing based in the positioning and tracing method of acoustic source array, the structural representation of the three-dimensional orientation geometric model that sound source and microphone array form;

Fig. 5 is provided by the present invention based in the positioning and tracing method of acoustic source array, the structural representation of the three-dimensional orientation geometric model that sound source and five yuan of microphone arrays form;

Fig. 6 is provided by the present invention based in the positioning and tracing method of acoustic source array, adopts broad sense cross correlation function method to calculate the process flow diagram in the relative time delay between every two-way audio signal.

Embodiment

Below in conjunction with the drawings and specific embodiments, technology contents of the present invention is described in further detail.

As shown in Figure 1, the positioning and tracking device based on acoustic source array provided by the present invention, comprises moving cell, control module, audio unit and display unit.Wherein, audio unit comprises audio frequency acquisition module, signal amplifier, filter module and A/D data acquisition module.Audio frequency acquisition module is in order to collection site sound, and signal amplifier is in order to amplify the sound signal of microphone collection each in multi-microphone array, and filter module is in order to carry out elementary filtering to sound signal.A/D data acquisition module in order to process sound signal do data conversion so that be transferred to control module do based on time delay estimate auditory localization computing.Sound signal after the process that control module audio reception unit sends, is processed by the computing of auditory localization algorithm, by process after auditory localization information deliver to display unit, meanwhile, by process after positioning control order be sent to moving cell.In embodiment provided by the present invention, the process of control module to audio-frequency information is completed by central processing unit.

Moving cell receives the control command from control module, and the positioning and tracking device being used for controlling acoustic source array carries out rotating and travelling according to orientation.Wherein, moving cell sends control signal to motor driver by control module, and drive motor completes corresponding go to action, makes the positioning and tracking device based on acoustic source array provided by the present invention towards the orientation of main sound source.Moving cell comprises two driving wheels, be arranged at the electric drive left and right sides respectively, complete advance or backward movement when two driving wheels synchronously move at the uniform velocity, when giving two driving wheel friction speeds, due to the existence of velocity contrast, positioning and tracking device is rotated.Display unit is completed by display, and display unit receives the state of auditory localization information and the positioning and tracking device obtained after treatment from control part unit, displays it.

Fig. 2 is the process flow diagram of the positioning and tracing method based on acoustic source array provided by the invention.First the method specifically comprises the steps:, adopts the on-the-spot sound of five yuan of microphone array acquisition, carries out pre-service obtain sound signal to the voice signal of microphone collection each in five yuan of microphone arrays.Then, carry out the location of sound signal according to the positional information of the sound signal time delay of arrival between each microphone and microphone array, calculate the angle of pitch, position angle and target range.Finally, according to the angle of pitch calculated, position angle and target range positioning and tracking device moved and turn to, sound source being positioned, and the state of auditory localization information and positioning and tracking device is shown.Detailed specific description is done to this process below.

S1, adopts the on-the-spot sound of five yuan of microphone array acquisition, carries out pre-service obtain sound signal to the voice signal of microphone collection each in five yuan of microphone arrays.

In existing quaternary microphone array, as shown in Figure 3, this array is respectively by microphone M ₁, M ₂, M ₃and M ₄composition, the three-dimensional orientation geometric model that voice signal and microphone array form as shown in Figure 4.The microphone array of cross formation picks up same voice signal, obtains the mistiming each other, then calculates the azimuth information of sound source according to the geometric relationship of cross formation.Planar four-element cruciform microphone array, can obtain three independently delay volumes, can obtain the angle of pitch of sound source in three dimensions, position angle and distance with this.

In embodiment provided by the present invention, adopt the on-the-spot sound of five yuan of microphone array acquisition.As shown in Figure 5, five microphone composition cross formations of five yuan of microphone arrays, pick up same voice signal.Wherein, five microphones forming five yuan of microphone arrays can in the optional position of cross formation.In embodiment provided by the present invention, the rectangular coordinate of each microphone position is respectively: S ₀(0,0,0), S ₁(D/2,0,0), S ₃(-D/2,0,0), S ₂(0, D/2,0), S ₄(0 ,-D/2,0), the rectangular coordinate of target sound source T is (x, y, z).D is the distance that microphone arrives true origin, then in coordinate axis, array element distance is D.Target is r to the distance of true origin, and position angle is the angle of pitch is θ.

Adopt the on-the-spot sound of five yuan of microphone array acquisition, pre-service is carried out to the voice signal of microphone collection each in five yuan of microphone arrays and obtains sound signal.Before carrying out pre-service to the voice signal gathered, do signal condition, exported and be amplified to applicable amplitude range by two-stage amplifying circuit, after A/D sampling, carry out the pre-service of voice signal, key step is as follows:

S11, carries out windowing framing by sampled signal (voice signal), obtains the signal segment of short-term stationarity.Wherein, framing is with specific window function w (n) and sampled signal s ₁n () carries out convolutional calculation, thus form windowing signal s _w(n)=s ₁(n) * w (n).

S12, carries out breaking point detection to sampled signal, calculates short-time energy E, if E is less than threshold value can think noise, wherein, n is sampling number, due to the calculating to relate to square, the little short-time average magnitude of calculated amount therefore also can be adopted to characterize the energy of sampled signal, namely otherwise, turn to step S13.

S13, calculates short-time zero-crossing rate Z ₀if, Z ₀in thresholding, then represent that voice signal is the sound signal of band noise.Short-time zero-crossing rate represents the number of times of waveform through transverse axis (zero level) of signal in a frame sampling signal, just means to discrete signal the number of times that this sampled signal changes.The windowing signal s of the sound signal with noise _wn the short-time zero-crossing rate of () is Z ₀: $Z_0=\frac{1}{2}\left\{\underset{n=1}{\overset{N-1}{\Σ}}\right|\mathrm{sgn}\[s_w\left(n\right)\]-\mathrm{sgn}\[s_w(n-1)\]\left|\right\};$

S14, after the sound signal of band noise detects, utilizes spectrum-subtraction denoising, sends into FIR bandpass filter and carries out filtering process.

After the sound signal of band noise detects, utilize spectrum-subtraction denoising, the basic thought of spectrum-subtraction is that supposition noise and sound signal are separate.Wherein, spectrum-subtraction denoising is utilized specifically to comprise the steps:

S141, carries out Fourier transform to the sound signal of band noise, obtains: X (ω)=S (ω)+N (ω), obtains power spectrum further: | X (ω) ²|=| S (ω) | ²+ | N (ω) | ²+ S (ω) N ^*(ω)+S ^*(ω) N (ω); Due to sound signal and noise separate, then S ^*(ω) N (ω) and S (ω) N ^*(ω) 0, P is _s(ω), P _x(ω) and P _n(ω) power spectrum of s (t), x (t) and n (t) is represented respectively, then P _x(ω)=P _s(ω)+P _n(ω), wherein, s (t) is sound signal, and n (t) is noise signal, and x (t) is the sound signal of band noise, x (t)=s (t)+n (t).

S142, gets the frame data before sound signal being detected as noise signal, obtains noise power spectrum P _n(ω);

S143, according to formula: P _s(ω)=P _x(ω)-P _n(ω) (namely from the power spectrum of the sound signal of band noise, deduct noise power spectrum) and obtain pure sound spectrum.Because people's ear is to the insensitivity of phase place, sound signal can be recovered by pure sound spectrum and band noise phase place frequently.

Get the frame data before sound signal being detected as noise signal, obtain its power spectrum P _n(ω), power spectrum signal P _x(ω) also can try to achieve, according to formula: P _s(ω)=P _x(ω)-P _n(ω) pure sound spectrum can be obtained.

S144, carries out to the pure sound spectrum obtained the amplitude spectrum that evolution obtains pure audio.

S145, makes Fourier transform to sound signal s (t), obtains its phase place, by this phase multiplication with the amplitude spectrum of pure audio, then carries out Fourier inversion and can obtain composing the sound signal after subtracting.In actual operation, for preventing the situation of negative power, when spectrum subtracts, work as P _x(ω) < P _n(ω), time, P is made _s(ω)=0.

When adopting spectrum-subtraction denoising, due to the projection of noise on some frequency, meeting residual fraction noise after subtracting each other, in embodiment provided by the present invention, remains by deducting some components to reduce more or eliminate this.Specifically can be realized by two kinds of modes, a kind of mode the noise power spectrum estimated is multiplied by the constant that is greater than 1, artificial increase noise power spectrum; Another kind of mode is the power spectrum signal after analytical spectra subtracts, and composes compare with original signal power, and the part being less than former power spectrum certain percentage (such as, original signal power 60%) is set to 0.

Sound signal after above-mentioned process, as the sound source estimated based on time delay, positions the computing of algorithm, thus obtains the orientation of main sound source.

S2, according to the positional information of the sound signal time delay of arrival between each microphone and microphone array to the location of carrying out pretreated sound signal and carry out sound source, calculates the angle of pitch, position angle and target range.

According to the positional information of the sound signal time delay of arrival between each microphone and microphone array to the location of carrying out pretreated sound signal and carry out sound source, specifically comprise the steps:

S21, by broad sense cross correlation function method, calculate five yuan of microphone array received to sound signal in, the relative time delay between every two-way audio signal.

Suppose that the coordinate of the discrete event signal model of two microphones voice signals is: x ₁(t)=a ₁s (t)+n ₁(t), x ₂(t)=a ₂s (t-τ ₁₂)+n ₂(t); Wherein, a ₁for the attenuation coefficient of sound-source signal, s (t) is sound signal, x _it voice signal that () gathers for microphone, n _it noise signal that () adds for audio frequency, τ ₁₂be the time delay of two microphone pickup sound signals, i.e. time delay.

In embodiment provided by the present invention, adopt broad sense cross correlation function method calculate five yuan of microphone array received to sound signal in, in the relative time delay between every two-way audio signal, as shown in Figure 6, specifically comprise the steps:

S211, asks Fourier transform (FFT) to two-way audio signal, gets conjugation to the data after a wherein road Fourier transform.In embodiment provided by the present invention, get microphone S respectively ₁, microphone S ₂, microphone S ₃, microphone S ₄with microphone S ₀between sound signal ask Fourier transform.

S212, calculates the cross-power spectrum between two-way audio signal, selects different weighting factors to be weighted cross-power spectrum, make the time delay place peak value of cross correlation function between two paths of signals more obvious according to different demands.

S213, is transformed into time domain by frequency-region signal, obtains broad sense cross correlation function.

Wherein, broad sense cross correlation function is: $R_{12}\left(\mathrm{\&tau;}\right)={\&Integral;}_0^{\mathrm{\&pi;}}{\mathrm{\&psi;}}_{12}\left(\mathrm{\&omega;}\right)X_1\left(\mathrm{\&omega;}\right)X_2^{*}\left(\mathrm{\&omega;}\right)e^{-j\mathrm{\&omega;}t}d\mathrm{\&omega;};$ X ₁(ω), X ₂(ω) be x respectively ₁(t), x ₂the Fourier transform of (t), it is cross-power spectrum.ψ ₁₂for weight function, in embodiment provided by the present invention, broad sense cross correlation function method is now called PHAT-GCC method, and the method is other broad sense cross correlation function method relatively, best to the rejection ability of reverberation.

S214, finds the peak of related function, is the relative time delay between sound signal.

Carry out peakvalue's checking, the point of the horizontal ordinate that the peak value of cross correlation function is corresponding is exactly time delay value t ₀₁, use the same method and can calculate time delay value t ₀₂, t ₀₃and t ₀₅, finally draw microphone S ₁, microphone S ₂, microphone S ₃with microphone S ₄with microphone S ₀between time delay value be respectively: t ₀₁, t ₀₂, t ₀₃, t ₀₄.

S215, filters the misdata in the relative time delay obtained.

The time delay calculated due to broad sense cross correlation function method is mingled with the situation of a large amount of misdata, and the relative time delay that therefore broad sense cross correlation function method estimates needs to do further process, thus filters misdata.In embodiment provided by the present invention, the effect that stochastic distribution screening reaches filter false data is carried out to multiframe time delay estimated result.Below the introduction of stochastic distribution filtering algorithm:

According to right value integrated distribution, the characteristic of improper value stochastic distribution, stochastic distribution filtering algorithm sorts to 10 frame time delay results, removes 3 maximal values and 3 minimum value, is averaged residue 4 values, and recycling average delay carries out orientation.Consider that the speech comparison that same person sends links up, get 3 seconds in one embodiment of the present of invention as critical point, if two sections of speech intervals were more than 3 seconds, then do not carried out combination screening.Specific implementation step is as follows:

S2151, carries out sound detection, detects that sound then enters next step, sound do not detected in the schedule time, then empty time delay value array; In embodiment provided by the present invention, the schedule time was set to for 3 seconds.

S2152, utilizes the acoustic segment comprised in time delay value array to calculate time delay value.

S21521, compares time delay number and delay threshold in time delay value array, when time delay number is less than delay threshold in time delay value array, the time delay value newly estimated is put into array end; In embodiment provided by the present invention, delay threshold gets 10.

S21522, when in time delay value array, time delay number equals delay threshold, then by individual to (delay threshold-1) after array time delay reach, then the time delay value calculated by new estimation puts into array end;

S21523, again judges time delay number in time delay value array, when being less than delay threshold, then turns to step S21521, when equaling delay threshold, then turn to step S21524;

S21524, sorts to the time delay value in time delay array, removes 3 maximal values and 3 minimum value, asks the average of residue (delay threshold-6) individual value, utilize this average delay to carry out orientation.

Above-mentioned stochastic distribution filtering algorithm by being averaging multiframe time delay, further refinement time delay value, thus have higher precision.This algorithm is comparatively applicable to sound and links up or time delay misjudgment is worth not many situation, verifying, achieving good effect by playing coherent sound with audio amplifier.

In addition, utilize broad sense cross correlation function method calculation delay, be actually the interval of first trying to achieve two digital signals and count, then be multiplied with signal sampling rate and try to achieve time delay.Visible samples rate will directly affect the precision of calculation delay, and sampling rate is higher, and the time delay precision estimated more is done.But it is limited for itself improving sampling rate by means of only hardware collection, and can raise the cost greatly and resource overhead.Wanting to obtain high-precision time delay value when reducing adding system expense, method of interpolation can be adopted to improve time-delay calculation precision.In embodiment provided by the present invention, the sound signal collected is carried out interpolation processing, improve sample frequency.

S22, according to the relative time delay that step S21 obtains, utilizes the geometric relationship of array element and sound source position to obtain the position of sound source.

Before address, in existing quaternary microphone array, array is respectively by microphone M ₁, M ₂, M ₃and M ₄composition, the microphone array of cross formation picks up same sound-source signal, obtains the mistiming each other, then calculates the azimuth information of sound source according to the geometric relationship of cross formation.

Suppose that the coordinate of four microphones in cross battle array is respectively:

M ₁(d, 0,0), M ₂(0, d, 0), M ₃(-d, 0,0), M ₄(0 ,-d, 0), the coordinate of sound source S in space is S (x, y, z), and sound source is r to the distance of true origin, and the angle of pitch is θ, and position angle is d is the distance that microphone arrives true origin, then in coordinate axis, array element distance is 2d.Suppose that sound source position is distant from microphone array, then have r>>2d.With first microphone M ₁for reference, the sound signal that sound source sends arrives M ₂, M ₃, M ₄time delay be designated as t respectively ₁₂, t ₁₃, t ₁₄, then in conjunction with velocity of sound c, can show that corresponding path difference is respectively:

$\left\{\begin{array}{c}{s}_{12}={\mathrm{ct}}_{12}\\ s_{13}={\mathrm{ct}}_{13}\\ s_{14}={\mathrm{ct}}_{14}\end{array}\right.---\left(3\right)$

In rectangular coordinate system, S (x, y, z) meets equation:

x ²+y ²+z ²＝r ²(4)

Then there is system of equations:

$\left\{\begin{array}{c}{(x-d)}^2+y^2+z^2={r_1}^2\\ x^2+{(y-d)}^2+z^2={(r_1+s_{12})}^2\\ {(x+d)}^2+y^2+z^2={(r_1+s_{13})}^2\\ x^2+{(y+d)}^2+z^2={(r_1+s_{14})}^2\end{array}\right.---\left(5\right)$

Wherein, sound-source signal is r to the distance of first microphone ₁, can be solved by system of equations:

$\{\begin{array}{c}x=\frac{2r_1{s}_{13}+s_{13}^2}{4d}\\ y=\frac{2r_1(s_{14}-s_{12})+s_{14}^2-s_{12}^2}{4d}\\ r_1=\frac{s_{12}^2+s_{14}^2-s_{13}^2}{2(s_{13}-s_{12}-s_{14})}\end{array}---\left(6\right)$

Sound-source signal S (x, y, z) be converted into polar coordinates be S (r, θ), corresponding relation is:

Wherein, 0 °≤θ≤90 °,

1) distance r is calculated:

$r=\frac{c(t_{12}^2+t_{14}^2-t_{13}^2)}{2(t_{13}-t_{12}-t_{14})}---\left(8\right)$

2) computer azimuth angle :

The position angle then asked:

3) pitching angle theta is calculated:

$\mathrm{\&theta;}=arcsin\frac{\sqrt{t_{13}^2+{(t_{14}-t_{12})}^2}\&CenterDot;c}{2d}---\left(11\right)$

In embodiment provided by the present invention, adopt the on-the-spot sound of five yuan of microphone array acquisition.Five microphone composition cross formations of five yuan of microphone arrays, pick up same sound-source signal.Wherein, the rectangular coordinate of each microphone position is respectively: S ₀(0,0,0), S ₁(D/2,0,0), S ₃(-D/2,0,0), S ₂(0, D/2,0), S ₄(0 ,-D/2,0), the rectangular coordinate of target sound source T is (x, y, z).D is the distance that microphone arrives true origin, then in coordinate axis, array element distance is D.Target is r to the distance of true origin, and position angle is the angle of pitch is θ.

Suppose that sound source T propagates with spherical wave form, arrive array element S ₁, S ₂, S ₃, S ₄relative to arrival array element S ₀time delay be respectively t ₀₁, t ₀₂, t ₀₃, t ₀₄, then in conjunction with the velocity of sound, can obtain its target localization equation is:

$\mathrm{\&theta;}\&ap;arcsin\left(\frac{C}{D}\sqrt{{(t_{01}-t_{03})}^2+{(t_{02}-t_{04})}^2}\right)---\left(13\right)$

$r=\frac{C^2({t_{01}}^2+{t_{02}}^2+{t_{03}}^2+{t_{04}}^2)-2D^2}{4C(t_{01}+t_{02}+t_{03}+t_{04})}---\left(14\right)$

Here be the time delay evaluated error of four array element cross battle arrays and five array element cross battle arrays to the impact analysis of the orientation accuracy of system:

First the Transfer Formula of error is as follows, utilizes this formula can derive the interact relation formula of time delay evaluated error to orientation accuracy.

${\mathrm{\&sigma;}}^2\left(y\right)=\underset{j=1}{\overset{m}{\&Sigma;}}{\left(\frac{\&part;f}{\&part;x_j}\right)}^2{\mathrm{\&sigma;}}^2\left(x_j\right)x_m---\left(15\right)$

Above formula is the variance composite formula when m item component is separate.

Four-element array azimuth accuracy is analyzed:

Due to time delay τ ₁₂, τ ₁₃, τ ₁₄statistical error feature identical, might as well τ be established _1ithe variance of (i=2,3,4) is δ _τ, then according to (15) formula, the azimuth angle error that can be caused by time delay evaluated error is:

$sin\mathrm{\&theta;}\&ap;\frac{C}{D}\sqrt{{({\mathrm{\&tau;}}_{12}-{\mathrm{\&tau;}}_{14})}^2+{{\mathrm{\&tau;}}_{13}}^2}---\left(18\right)$

Ask partial derivative respectively:

Partial derivative (16) formula of bringing into (17) formula is asked to obtain:

The precision analysis of the four-element array angle of pitch:

For pitching angle theta, catch up with to state and similarly obtain formula:

${\mathrm{\&delta;}}_{\mathrm{\&theta;}\mathrm{\&tau;}}=\sqrt{{\left(\frac{\&part;\mathrm{\&theta;}}{\&part;{\mathrm{\&tau;}}_{12}}{\mathrm{\&delta;}}_{\mathrm{\&tau;}}\right)}^2+{\left(\frac{\&part;\mathrm{\&theta;}}{\&part;{\mathrm{\&tau;}}_{13}}{\mathrm{\&delta;}}_{\mathrm{\&tau;}}\right)}^2+{\left(\frac{\&part;\mathrm{\&theta;}}{\&part;{\mathrm{\&tau;}}_{14}}{\mathrm{\&delta;}}_{\mathrm{\&tau;}}\right)}^2}$

Partial derivative is asked to obtain to formula (18) formula:

Distance estimations precision analysis:

Target range

$r\&ap;r_1=\frac{C}{2}\&CenterDot;\frac{({{\mathrm{\&tau;}}_{12}}^2+{{\mathrm{\&tau;}}_{14}}^2-{{\mathrm{\&tau;}}_{13}}^2)}{({\mathrm{\&tau;}}_{13}-{\mathrm{\&tau;}}_{12}-{\mathrm{\&tau;}}_{14})},$ In conjunction with with $sin\mathrm{\&theta;}=\frac{C}{D}\sqrt{{({\mathrm{\&tau;}}_{12}-{\mathrm{\&tau;}}_{14})}^2+{{\mathrm{\&tau;}}_{13}}^2}.$

The root-mean-square error of Target Distance Estimation is:

Consider that the relative error of Target Distance Estimation is:

Therefore can find out, time delay estimate to cause azimuth angle error, angle of pitch error, target range error all with velocity of sound C, array element distance D, pitching angle theta and position angle relevant.

The Accuracy Analysis of five element array:

Azimuth accuracy is analyzed:

If τ _0ithe variance of (i=1,2,3,4) is δ _τ, then time delay estimates that the azimuth angle error caused is:

Respectively partial derivative is asked to above formula:

By above formula simultaneous, substitute into (23) formula:

By finding out, for five array element cross battle arrays, time delay estimates that the azimuth angle error caused is relevant with pitching angle theta with velocity of sound C, array element distance D, and follows the position angle of target sound source irrelevant.When Bian is by four array element cross battle arrays, array element number reduces, and reduces the complicacy of system, but when azimuth of target θ is near 0 degree or 180 degree, position angle the error estimated is larger.And the orientation accuracy of five array element cross array 1 systems and the position angle of target irrelevant.

The precision that in five yuan of battle arrays, position angle is estimated is relevant with the angle of pitch of array element distance and target, and has nothing to do with the position angle of target, the shortcoming affected by azimuth of target when overcoming the battle array direction finding of quaternary cross.For given Time delay Estimation Accuracy, with the increase of array element distance, position angle estimated accuracy improves; And target pitch angle is larger, the root-mean-square error that position angle is estimated is less, and precision is also higher.

Angle of pitch precision analysis:

In like manner, according to $\sin \mathrm{\&theta;}\&ap;\frac{C}{D}\sqrt{{({\mathrm{\&tau;}}_{01}-{\mathrm{\&tau;}}_{03})}^2+{({\mathrm{\&tau;}}_{02}-{\mathrm{\&tau;}}_{04})}^2},$ Can try to achieve the angle of pitch to the partial derivative of each time delay is:

$\left\{\begin{array}{c}\frac{\&part;\mathrm{\&theta;}}{\&part;{\mathrm{\&tau;}}_{01}}=-\frac{\&part;\mathrm{\&theta;}}{\&part;{\mathrm{\&tau;}}_{03}}\&ap;\frac{2C^2}{D^{2}sin2\mathrm{\&theta;}}({\mathrm{\&tau;}}_{01}-{\mathrm{\&tau;}}_{03})\\ \frac{\&part;\mathrm{\&theta;}}{\&part;{\mathrm{\&tau;}}_{02}}=-\frac{\&part;\mathrm{\&theta;}}{\&part;{\mathrm{\&tau;}}_{04}}\&ap;\frac{2C^2}{D^{2}sin2\mathrm{\&theta;}}({\mathrm{\&tau;}}_{02}-{\mathrm{\&tau;}}_{04})\end{array}\right.$

Therefore angle of pitch variance can be expressed as with the relational expression between time delay variance:

$\begin{array}{c}{\mathrm{\&delta;}}_{\mathrm{\&theta;}\mathrm{\&tau;}}=\sqrt{{\left(\frac{\&part;\mathrm{\&theta;}}{\&part;{\mathrm{\&tau;}}_{01}}{\mathrm{\&delta;}}_{\mathrm{\&tau;}}\right)}^2+{\left(\frac{\&part;\mathrm{\&theta;}}{\&part;{\mathrm{\&tau;}}_{02}}{\mathrm{\&delta;}}_{\mathrm{\&tau;}}\right)}^2+{\left(\frac{\&part;\mathrm{\&theta;}}{\&part;{\mathrm{\&tau;}}_{03}}{\mathrm{\&delta;}}_{\mathrm{\&tau;}}\right)}^2+{\left(\frac{\&part;\mathrm{\&theta;}}{\&part;{\mathrm{\&tau;}}_{04}}{\mathrm{\&delta;}}_{\mathrm{\&tau;}}\right)}^2}\\ =\frac{2\sqrt{2}C}{D\cos \mathrm{\&theta;}}{\mathrm{\&delta;}}_{\mathrm{\&tau;}}\end{array}$

The estimated accuracy at target pitch angle is relevant with array element distance and target pitch angle, and has nothing to do with position angle.

Distance estimations precision analysis:

Can be obtained the partial derivative of each time delay by distance r:

$\frac{\&part;r}{\&part;{\mathrm{\&tau;}}_{0i}}\&ap;\frac{2rC({\mathrm{C\&tau;}}_{0i}-r)}{D^2({\sin }^2\mathrm{\&theta;}-4)},(i=1,2,3,4)$

Therefore distance variance can be expressed as with the relational expression between time delay variance:

$\begin{array}{c}{\mathrm{\&delta;}}_{r\mathrm{\&tau;}}=\sqrt{{\left(\frac{\&part;r}{\&part;{\mathrm{\&tau;}}_{01}}{\mathrm{\&delta;}}_{\mathrm{\&tau;}}\right)}^2+{\left(\frac{\&part;r}{\&part;{\mathrm{\&tau;}}_{02}}{\mathrm{\&delta;}}_{\mathrm{\&tau;}}\right)}^2+{\left(\frac{\&part;r}{\&part;{\mathrm{\&tau;}}_{03}}{\mathrm{\&delta;}}_{\mathrm{\&tau;}}\right)}^2+{\left(\frac{\&part;r}{\&part;{\mathrm{\&tau;}}_{04}}{\mathrm{\&delta;}}_{\mathrm{\&tau;}}\right)}^2}\\ \&ap;\frac{4rC\sqrt{D^2+r^2}}{(4-{\sin }^2\mathrm{\&theta;})}{\mathrm{\&delta;}}_{\mathrm{\&tau;}}\end{array}$

The positioning precision of target range is relevant with the angle of pitch of array element distance, target range, target, and has nothing to do with the position angle of target.In time delay evaluated error, array element distance and target range one timing, Target Distance Estimation precision is relevant with the angle of pitch, and has nothing to do with position angle.Target pitch angle larger, positioning precision is higher.

S3, to move positioning and tracking device according to the angle of pitch calculated, position angle and target range and turns to, and arrives sound source position.

By two driving wheels being arranged at the electric drive left and right sides, positioning and tracking device is turned to according to the angle of pitch calculated, position angle and relative distance, sound source is positioned, and the state of auditory localization information and positioning and tracking device is shown.

In sum, the positioning and tracing method based on acoustic source array provided by the present invention, adopts the on-the-spot sound of five yuan of microphone array acquisition, carries out pre-service obtain sound signal to the voice signal of microphone collection each in five yuan of microphone arrays.Carry out the location of sound source according to the positional information of the sound signal time delay of arrival between each microphone and microphone array, calculate the angle of pitch, position angle and target range.Finally, according to the angle of pitch calculated, position angle and target range positioning and tracking device moved and turn to, localization of sound source, and the state of auditory localization information and positioning and tracking device is shown.Reverberation for non-Gaussian noise, coherent noise and sound source originally in indoor is aimed at and is determined the impact that position is caused, and has carried out relevant treatment, has improve the accuracy of auditory localization.

Above the positioning and tracing method based on acoustic source array provided by the present invention and system thereof are described in detail.For one of ordinary skill in the art, to any apparent change that it does under the prerequisite not deviating from connotation of the present invention, all by formation to infringement of patent right of the present invention, corresponding legal liabilities will be born.

Claims (10)

1., based on a positioning and tracing method for acoustic source array, it is characterized in that comprising the steps:

S1, adopts the on-the-spot sound of five yuan of microphone array acquisition, carries out pre-service obtain sound signal to the voice signal of microphone collection each in five yuan of microphone arrays;

S2, carries out the location of sound source, calculates the angle of pitch, position angle and target range to sound signal according to the positional information of the sound signal time delay of arrival between each microphone and microphone array;

S3, to move positioning and tracking device according to the angle of pitch calculated, position angle and target range and turns to, and arrives sound source position.

2., as claimed in claim 1 based on the location moving method of acoustic source array, it is characterized in that:

In step sl, five microphone composition cross formations in described five yuan of microphone arrays, for picking up same sound-source signal.

3., as claimed in claim 1 based on the positioning and tracing method of acoustic source array, it is characterized in that in step sl, voice signal carries out pre-service and obtains sound signal, comprises the steps:

S11, carries out windowing framing by voice signal, obtains the signal segment of short-term stationarity; Wherein, framing is with specific window function w (n) and voice signal s ₁n () carries out convolutional calculation, thus form windowing signal s _w(n)=s ₁(n) * w (n);

S12, carries out breaking point detection to voice signal, calculates short-time energy E, if E is less than threshold value can think noise, wherein, n is sampling number, otherwise, turn to step S13;

S13, calculates short-time zero-crossing rate Z ₀if, Z ₀in thresholding, then represent sound signal band noise being detected, the windowing signal s of the sound signal of band noise _wn the short-time zero-crossing rate of () is Z ₀: $Z_0=\frac{1}{2}\left\{\underset{n=1}{\overset{N-1}{\&Sigma;}}\right|\mathrm{sgn}\&lsqb;s_w\left(n\right)\&rsqb;-\mathrm{sgn}\&lsqb;s_w(n-1)\&rsqb;\left|\right\};$

S14, after the sound signal of band noise detects, utilizes spectrum-subtraction denoising, and feeding FIR bandpass filter is carried out filtering process and obtained sound signal.

4., as claimed in claim 3 based on the positioning and tracing method of acoustic source array, it is characterized in that in step S14, utilize spectrum-subtraction denoising, comprise the steps:

S141, carry out Fourier transform to the sound signal of band noise, obtain: X (ω)=S (ω)+N (ω), rated output is composed: | X (ω) ²|=| S (ω) | ²+ | N (ω) | ²+ S (ω) N ^*(ω)+S ^*(ω) N (ω); Wherein, S ^*(ω) N (ω) and S (ω) N ^*(ω) 0, P is _s(ω), P _x(ω) and P _n(ω) represent the power spectrum of s (t), x (t) and n (t) respectively, s (t) is sound signal, and n (t) is noise signal, and x (t) to make an uproar sound signal, then P for band _x(ω)=P _s(ω)+P _n(ω);

S142, gets the frame data before sound signal being detected as noise signal, obtains noise power spectrum P _n(ω);

S143, according to formula: P _x(ω)=P _s(ω)+P _n(ω) pure sound spectrum is obtained;

S144, carries out to the pure sound spectrum obtained the amplitude spectrum that evolution obtains pure audio;

S145, makes Fourier transform to sound signal s (t), obtains its phase place, by described phase multiplication with the amplitude spectrum of pure audio, then carries out Fourier inversion and obtains composing the sound signal after subtracting.

5. as claimed in claim 1 based on the positioning and tracing method of acoustic source array, it is characterized in that in step s 2, according to the positional information of the sound signal time delay of arrival between each microphone and microphone array to the location of carrying out pretreated sound signal and carry out target sound source, comprise the steps:

S21, by broad sense cross correlation function method, calculate five yuan of microphone array received to sound signal in, the relative time delay between every two-way audio signal;

S22, according to the relative time delay that step S21 obtains, utilizes the geometric relationship of array element and sound source position to obtain the position of sound source;

The position of described sound source comprises the position angle of source of students, the angle of pitch and target range.

6. as claimed in claim 5 based on the positioning and tracing method of acoustic source array, it is characterized in that in the step s 21, adopt broad sense cross correlation function method calculate five yuan of microphone array received to sound signal in, in the relative time delay between every two-way audio signal, comprise the steps:

S211, asks Fourier transform to two-way audio signal, gets conjugation to the data after a wherein road Fourier transform;

S212, calculates the cross-power spectrum between two-way audio signal;

S213, is transformed into time domain by frequency-region signal, obtains broad sense cross correlation function, and wherein, broad sense cross correlation function is: $R_{12}\left(\mathrm{\&tau;}\right)={\&Integral;}_0^{\mathrm{\&pi;}}{\mathrm{\&psi;}}_{12}\left(\mathrm{\&omega;}\right)X_1\left(\mathrm{\&omega;}\right)X_2^{*}\left(\mathrm{\&omega;}\right)e^{-j\mathrm{\&omega;}t}d\mathrm{\&omega;};$ X ₁(ω), X ₂(ω) be x respectively ₁(t), x ₂the Fourier transform of (t), cross-power spectrum, ψ ₁₂for weight function;

S214, finds the peak of broad sense cross-correlation related function, is the relative time delay between sound signal;

S215, filters the misdata in the relative time delay obtained.

7., as claimed in claim 6 based on the positioning and tracing method of acoustic source array, it is characterized in that in step S215, the misdata in the relative time delay obtained is filtered, comprises the steps:

S2151, carries out sound detection, detects that sound then enters next step, sound do not detected in the schedule time, then empty time delay value array;

S2152, utilizes the acoustic segment comprised in time delay value array to calculate time delay value.

8., as claimed in claim 7 based on the positioning and tracing method of acoustic source array, it is characterized in that in step S2152, calculate time delay value and comprise the steps:

S21521, compares time delay number and delay threshold in time delay value array, when time delay number is less than delay threshold in time delay value array, the time delay value newly estimated is put into array end;

S21522, when in time delay value array, time delay number equals delay threshold, then by-1 the time delay reach of delay threshold after array, then the time delay value calculated by new estimation puts into array end;

S21523, again judges time delay number in time delay value array, when being less than delay threshold, then turns to step S21521, when equaling delay threshold, then turn to step S21524;

S21524, sorts to the time delay value in time delay array, removes 3 maximal values and 3 minimum value, asks the average of residual time delay threshold value-6 values, this average delay is carried out the orientation of sound source as time delay value.

9., based on a positioning and tracking device for acoustic source array, for realizing the positioning and tracing method based on acoustic source array according to claim 1, it is characterized in that comprising moving cell, control module, audio unit and display unit;

Wherein, described audio unit in order to collection site sound, and carries out pre-service to the voice signal of microphone collection each in five yuan of microphone arrays;

Sound signal after the process that described control module audio reception unit sends, is processed by auditory localization algorithm, and the auditory localization information after process is delivered to display unit, meanwhile, the positioning control order after process is sent to moving cell;

Described moving cell receives the control command from control module, and the positioning and tracking device being used for controlling acoustic source array carries out rotating and travelling according to orientation;

Described display unit receives the state of auditory localization information and the positioning and tracking device obtained after treatment from control part unit, displays it.

10., as claimed in claim 9 based on the positioning and tracking device of acoustic source array, it is characterized in that:

Described moving cell sends control signal to motor driver by control module, and drive motor completes corresponding go to action, makes positioning and tracking device based on acoustic source array towards the orientation of main sound source;

Described moving cell comprises two driving wheels, is arranged at the left and right sides of electric drive respectively, completes advance or backward movement when two driving wheels synchronously move at the uniform velocity, and when giving two driving wheel friction speeds, positioning and tracking device is rotated.