CN101902674A - Self-excited cancellation method for high-gain sound reinforcement system based on spatial cancellation - Google Patents

Abstract

The invention discloses a self-excited elimination method for a high-gain sound amplification system based on space cancellation, which performs weighted processing on the amplified voice signals of each frequency component and outputs them from two sound box speakers, so that the two output signals cancel each other at the microphone. In this way, by preventing the sound output by the speaker from feeding back to the microphone, the microphone loop can have extremely high amplification gain. This method makes the two self-excited signals of the feedback self-cancellation, independent of the target signal, does not process the target signal, and does not generate additional suppression signals; on the other hand, it eliminates the signal that causes the howling by itself, while Instead of filtering out the howling signal from the target signal, this fundamentally solves the self-excitation problem of the microphone.

Description

High-gain public address system self-excitation eliminating method based on the space counteracting

Technical field

The present invention relates to self-excitation eliminating method, specifically, relate to a kind of high-gain public address system self-excitation eliminating method of offsetting based on the space based on the space counteracting.

Background technology

When using public address system that sound is amplified in the actual life, source of sound must be enough near from microphone, otherwise just need transfer the gain of microphone big.When the gain of microphone was amplified to a certain degree, system will utter long and high-pitched sounds, and this is because of the raising along with gain amplifier, the probability (sensitivity) that microphone picks up noise has improved, get back to microphone through the sound of loudspeaker amplification and amplified once more, thereby form self-excitation, this self-excitation is uttered long and high-pitched sounds exactly.

The existing inhibition method of uttering long and high-pitched sounds has random phase compensation technique and the power spectrum technology etc. that disappears mutually.

Publication number is that the patent documentation 1 of CN1767695 has proposed a kind of " the inhibition method of uttering long and high-pitched sounds of nondestructive voice quality ", it is characterized in that at first with certain audio input signal constantly by analysis bank of filters obtain M sub-tape input signal; Then to the judgement of uttering long and high-pitched sounds of M/2+1 wherein subband input signal; If judging to exist in certain subband input signal utters long and high-pitched sounds, then its corresponding subband output signal obtains through a random phase device, does not utter long and high-pitched sounds if do not exist, and then it is the subband output signal; With M/2+1 the subband output signal that above-mentioned steps obtains, utilize the conjugate property of subband signal to obtain whole M subband output signals; With the above-mentioned M that obtains a subband output signal process composite filter group, obtain audio output signal at last.

The defective of patent documentation 1 is that it is that the audio signal of feedback is disappeared mutually with the audio signal of itself.Such shortcoming is to have weakened on the one hand to want the useful signal that amplifies, and it comes down to solve the self-excitation problem by reducing system gain on the other hand, and it has still limited the gain amplifier of system.

Publication number is that the patent documentation 2 of CN1398054 has proposed a kind of " utter long and high-pitched sounds detection and suppression equipment, method and computer program product ", its feature comprises: the frequency decomposition processing section is used for each sound time signal section corresponding to a period is converted to each sound frequency signal segment corresponding to a frequency range; The inhibition part of uttering long and high-pitched sounds, the gain of the sound frequency signal segment that the resolution process that is used for adjusting frequency is respectively partly changed generates the sound-inhibiting frequency signal section of uttering long and high-pitched sounds; The test section of uttering long and high-pitched sounds, be used to judge each suppressing portion of uttering long and high-pitched sounds divides the sound-inhibiting frequency signal section of uttering long and high-pitched sounds that generates whether to have the whistle composition, detects the no howl frequency signal segment that wherein has the howl frequency signal segment of whistle composition and wherein do not have the whistle composition; With the frequency synthesis processing section, be used for synthetic uttering long and high-pitched sounds and suppress the sound-inhibiting frequency signal section of uttering long and high-pitched sounds that part suppresses, the generation sound-inhibiting time signal section of uttering long and high-pitched sounds, thereby, utter long and high-pitched sounds suppress part by changing the howl frequency signal segment gain and allow no howl frequency signal segment pass through, adjust the gain of sound frequency signal segment respectively.Frequency decomposition processing section wherein, the method that makes need be a processing time unit with a period.It need generate the extra inhibition signal of uttering long and high-pitched sounds in addition, and this is similar to " the inhibition method of uttering long and high-pitched sounds of nondestructive voice quality ", and what difference was its adjustment is the detected gain that has the howl frequency signal segment of whistle composition.Simultaneously, patent documentation 1 and 2 has a common defective to be that they all are that the audio signal of feedback is disappeared mutually with the audio signal of itself.Such shortcoming is to have weakened on the one hand to want the useful signal that amplifies, and they come down to solve the self-excitation problem by reducing system gain on the other hand, so they have still limited the gain amplifier of system, can only suppress to utter long and high-pitched sounds but can not eliminate and utter long and high-pitched sounds.

Publication number is characterized in that producing corresponding the 1st power spectrum of the 1st acoustic signal that receives and export with described the 1st microphone for the patent documentation 3 of CN 1926911A provides a kind of " restraining device of uttering long and high-pitched sounds, program, integrated circuit and the inhibition method of uttering long and high-pitched sounds ".Produce with relate to comprise described expansion sound at least and do not comprise described purpose sound sound with corresponding the 2nd power spectrum of the 2nd acoustic signal.Then, according to described the 1st power spectrum and described the 2nd power spectrum, described the 1st acoustic signal of filtering, and the acoustic signal that only will be referred to described purpose sound outputs to the voice amplifier section.

The feature request of patent documentation 3 obtains to comprise expansion sound at least and the power spectrum that do not comprise purpose sound, and this feature makes the complexity of system increase on the one hand, needs the delay of voice signal on the other hand.

Summary of the invention

The object of the present invention is to provide a kind of high-gain public address system self-excitation eliminating method of offsetting, can solve the shortcoming of each method of background technology based on the space.

For reaching above purpose, the present invention takes following technical scheme to be achieved:

A kind of high-gain public address system self-excitation eliminating method of offsetting based on the space is characterized in that, may further comprise the steps:

(1) debug phase

A. processing module control A and B audio amplifier send a stable N simple signal: x ₁..., x _i..., x _N

B. to each simple signal: the one tunnel directly merges the back without any processing is exported by the A audio amplifier; Another road each frequency independence weighted amplitude and phase place obtain N the simple signal x after the processing ₁' ..., x _i' ..., x _N', wherein

A _iBe the amplitude weight coefficient

Be the phase place weight coefficient; Find the best weights coefficient

Make the signal x after the processing _i' with the A audio amplifier in corresponding simple signal x _iDisappear mutually at the microphone place later on is zero;

(2) stage is eliminated in self-excitation

D) the fixing debug phase best weights coefficient seeking out;

E) be N narrow band signal after the voice broadband signal process frequency slice from microphone, with the N in the debug phase simple signal x ₁..., x _i..., x _NBe centre frequency, this narrow band signal is divided into two-way;

F) the one tunnel directly merges and is exported by the A audio amplifier; Another road is by separately independently behind the weighting filter, and stack merges and exported by the B audio amplifier again.

In the such scheme, described searching best weights coefficient

Method be:

The first step is sought phase place

If the simple signal that the A audio amplifier sends is The simple signal that the B audio amplifier sends is Wherein

Be the phase place that spatial transmission causes, A ₁, A ₂Be the amplitude of two signals,

Phase place for weighting; In one-period, adopt 256 points, calculate the quadratic sum of each point amplitude, be designated as emp1, derivation can get emp1 with

Be sinusoidal rule and change, the relation between them can be expressed as

A wherein, B, C are three unknown parameters undetermined; Get four phase places

Write down corresponding emp1 ₁, emp1 ₂, emp1 ₃, emp1 ₄, obtain A by least square method, B, C.

Note

ε is a white Gaussian noise in the formula,

$Y=\left[\begin{array}{c}{\mathrm{<figure-callout\; id="1"\; label="emp"\; filenames="HDA0000024458400000021.png"\; state="\{\{state\}\}">emp</figure-callout>}1}_1\\ {\mathrm{<figure-callout\; id="1"\; label="emp"\; filenames="HDA0000024458400000021.png"\; state="\{\{state\}\}">emp</figure-callout>}1}_2\\ {\mathrm{<figure-callout\; id="1"\; label="emp"\; filenames="HDA0000024458400000021.png"\; state="\{\{state\}\}">emp</figure-callout>}1}_3\\ {\mathrm{<figure-callout\; id="1"\; label="emp"\; filenames="HDA0000024458400000021.png"\; state="\{\{state\}\}">emp</figure-callout>}1}_4\end{array}\right],$

$Z=\left[\begin{array}{c}C\\ A\\ B\end{array}\right],$

Z=(X then ^TX) ^-1X ^TY obtains A, B, C; And then try to achieve

For:

In second step, seek amplitude A _i

If the simple signal that the A audio amplifier that microphone is received sends is The B audio amplifier is

In one-period, adopt 256 points, calculate the quadratic sum of each point amplitude, be designated as emp2, A ₁Remain unchanged, change A ₂, derivation can get emp2 with A ₂Be parabola rule and change, then emp2=aA ₂ ²+ bA ₂+ c, a wherein, b, c are three unknown parameters undetermined.

Get four range value A ₂₁, A ₂₂, A ₂₃, A ₂₄, write down corresponding emp2 ₁, emp2 ₂, emp2 ₃, emp2 ₄, obtain undetermined parameter a, b, c by least square method.

Note emp2=aA ₂ ²+ bA ₂+ c+ ε, and ε～N (0, σ ²), ε is a white Gaussian noise in the formula,

$Q=\left[\begin{array}{c}{\mathrm{emp}2}_1\\ {\mathrm{emp}2}_2\\ {\mathrm{emp}2}_3\\ {\mathrm{emp}2}_4\end{array}\right],$ $P=\left[\begin{array}{ccc}1& {A_{21}}^2& A_{21}\\ 1& {A_{22}}^2& A_{22}\\ 1& {A_{23}}^2& A_{23}\\ 1& {A_{24}}^2& A_{24}\end{array}\right],$ $K=\left[\begin{array}{c}c\\ a\\ b\end{array}\right],$

Then can get K=(P ^TP) ^-1P ^TQ, and then the amplitude of trying to achieve is

The present invention adopts the space to offset mode, and the voice signal difference weighted of amplifying each frequency component of back is exported from the two-way speaker, and the two-way output signal is cancelled out each other at the microphone place.Like this, return the method for microphone, make the microphone loop can have high gain amplifier by the sound feedback that stops audio amplifier output.

The present invention disappears the two-way self-excitation signal of feedback voluntarily mutually, and is independent with the purpose signal, the purpose signal do not handled, and also do not generate extra inhibition signal simultaneously; On the other hand, it is that the signal that causes uttering long and high-pitched sounds is eliminated voluntarily, rather than from the purpose signal filtering signal of uttering long and high-pitched sounds, this has just fundamentally solved the self-excitation problem of microphone, the microphone that the obtains a kind of high-gain removing method of uttering long and high-pitched sounds.

Simultaneously, we have only adopted N level resonator here, but be equivalent to the voice broadband signal has been carried out the section of M=2* (N+1) section, so the frequency slice method among the present invention are a kind of frequency slice methods that effectively reduces hardware spending.Because, for a positive integer M,

So the i rank of frequency sampling mode filter and the output of M-i rank resonator are the conjugation symmetries, equal 2 times of real part of i rank output after their merge, and remove direct current and high frequency in actual applications, so, when practical application, the frequency sampling mode filter only needs to adopt The rank resonator is got real part after resonator output of each rank merges then as y (n).

Because people's ear is insensitive to phase place, and this method mainly changes the phase place of output sound signal, so be a kind of removing method of uttering long and high-pitched sounds of almost nondestructive voice quality to the processing of each narrow band signal.Simultaneously, it is independent of microphone and audio amplifier, thereby can be applied in various types of sound reinforcement systems, has great using value.

Description of drawings

Fig. 1 is the principle schematic of the inventive method.

Fig. 2 is the concrete structure figure of processing module among Fig. 1.

Fig. 3 is resonator H among Fig. 2 _k(z) concrete structure.

Embodiment

The present invention is described in detail by concrete enforcement below in conjunction with accompanying drawing.

As shown in Figure 1, microphone input signal is handled the back by two audio amplifier A, B outputs by processing module, cancels out each other at the microphone place (representing with fork).The processing module handling principle as shown in Figure 2, processing module mainly comprises: AD conversion, frequency slice part, weighting adjustment member, stack assembling section and DA conversion back output.Concrete implementation step is as follows:

The present invention's removing method of uttering long and high-pitched sounds needs to debug with simple signal earlier before application.This moment is the position of microphone and audio amplifier fixedly.

(1) debug phase

A. processing module control A and B audio amplifier send a stable N simple signal: x ₁..., x _i..., x _N

B. to each simple signal: the one tunnel directly merges the back without any processing is exported by the A audio amplifier; Each frequency of another road multiply by by weighting filter

The amplitude of carrying out and phase weighting obtain N the simple signal x after the processing ₁' ..., x _i' ..., x _N';

A _iBe the amplitude weight coefficient

Be the phase place weight coefficient; Find the weight coefficient that is suitable for current actual environment most

Make the signal x after the processing _i' with the A audio amplifier in corresponding simple signal x _iDisappear mutually at the microphone place later on is zero;

The best weights coefficient

Acquisition be divided into for two steps:

The first step is sought phase place

If the simple signal that the A audio amplifier sends is

The simple signal that the B audio amplifier sends is

Wherein

Be the phase place that spatial transmission causes, A ₁, A ₂Be the amplitude of two signals,

Phase place for weighting; In one-period, adopt 256 points, calculate the quadratic sum of each point amplitude, be designated as emp1, derivation can get emp1 with

Be sinusoidal rule and change, the relation between them can be expressed as

A wherein, B, C are three unknown parameters undetermined; Get four phase places

Write down corresponding emp1 ₁, emp1 ₂, emp1 ₃, emp1 ₄, obtain the ideal curve equation by least square method

ε is a white Gaussian noise in the formula,

Note $Y=\left[\begin{array}{c}{\mathrm{<figure-callout\; id="1"\; label="emp"\; filenames="HDA0000024458400000021.png"\; state="\{\{state\}\}">emp</figure-callout>}1}_1\\ {\mathrm{<figure-callout\; id="1"\; label="emp"\; filenames="HDA0000024458400000021.png"\; state="\{\{state\}\}">emp</figure-callout>}1}_2\\ {\mathrm{<figure-callout\; id="1"\; label="emp"\; filenames="HDA0000024458400000021.png"\; state="\{\{state\}\}">emp</figure-callout>}1}_3\\ {\mathrm{<figure-callout\; id="1"\; label="emp"\; filenames="HDA0000024458400000021.png"\; state="\{\{state\}\}">emp</figure-callout>}1}_4\end{array}\right],$

$Z=\left[\begin{array}{c}C\\ A\\ B\end{array}\right],$

Z=(X then ^TX) ^-1X ^TY obtains A, B, C; In order to make the emp1 minimum,

Determine by following formula:

In second step, seek amplitude A _i

If the simple signal that the A audio amplifier that microphone is received sends is

The B audio amplifier is

In one-period, adopt 256 points, calculate the quadratic sum of each point amplitude, be designated as emp2, A ₁Remain unchanged, change A ₂, derivation can get emp2 with A ₂Be parabola rule and change, then emp2=aA ₂ ²+ bA ₂+ c, a wherein, b, c are three unknown parameters undetermined.

Get four range value A ₂₁, A ₂₂, A ₂₃, A ₂₄, write down corresponding emp2 ₁, emp2 ₂, emp2 ₃, emp2 ₄, obtain undetermined parameter by least square method, emp2=aA ₂ ²+ bA ₂+ c+ ε, and ε～N (0, σ ²), ε is a white Gaussian noise in the formula,

Note $Q=\left[\begin{array}{c}{\mathrm{emp}2}_1\\ {\mathrm{emp}2}_2\\ {\mathrm{emp}2}_3\\ {\mathrm{emp}2}_4\end{array}\right],$ $P=\left[\begin{array}{ccc}1& {A_{21}}^2& A_{21}\\ 1& {A_{22}}^2& A_{22}\\ 1& {A_{23}}^2& A_{23}\\ 1& {A_{24}}^2& A_{24}\end{array}\right],$ $K=\left[\begin{array}{c}c\\ a\\ b\end{array}\right],$

Then can get K=(P ^TP) ^-1P ^TQ, and then try to achieve phase place and be

Signal to N frequency takes turns doing similar processing, obtains the best weights coefficient

The self-excitation elimination stage in the back uses.

(2) stage is eliminated in self-excitation

In actual applications, the signal gathered of microphone is a wideband speech signal.

G) the fixing debug phase best weights coefficient seeking out;

H) voice signal is realized frequency slice through overfrequency selection type filter.Frequency selection type Filter Structures is as shown in Fig. 2 " frequency slice " square frame, and it is to be formed by two parts cascade, and first is a comb filter of being made up of N joint delay unit, and second portion is the resonator of one group of parallel connection.The concrete structure of each resonator is made up of two multipliers, an adder and a delay unit, wherein as shown in Figure 3

R is one and is slightly less than 1 arithmetic number, is in order to guarantee the frequency slice system stability.

After the voice broadband signal process frequency slice from microphone is N narrow band signal, with the N in the debug phase simple signal x ₁..., x _i..., x _NBe centre frequency, this narrow band signal is divided into two-way;

One the tunnel directly merges and is exported by the A audio amplifier; Another road is weighted the back through weighting filter and merges by the output of B audio amplifier, and the structure of weighting filter is as shown in Fig. 2 " weighting " square frame, by a multiplier to amplitude and phase weighting.

Claims (2)

1. A self-excited cancellation method for a high-gain public address system based on space cancellation, characterized in that it may further comprise the steps: (1) Debugging stage a. The processing module controls speakers A and B to send out stable N single-frequency signals: x ₁ ,..., _xi ,...,x _N ; b. For each single-frequency signal: one channel is directly combined without any processing and then output by speaker A; the other channel independently weights the amplitude and phase of each frequency to obtain N processed single-frequency signals x ₁ ', ..., x _i ' ,..., x _N ', where

A _i is the amplitude weight coefficient

is the phase weight coefficient; find the optimal weight coefficient

Make the processed signal x _i ' and the corresponding single-frequency signal x _i in the A speaker cancel to zero at the microphone; (2) Self-excited elimination stage a) The best weight coefficient found in the fixed debugging stage; b) After frequency slicing, the voice broadband signal from the microphone is divided into N narrowband signals, and the N single-frequency signals x ₁ ,..., _xi ,...,x _N in the debugging stage are used as the center frequency, and the narrowband signal is divided into Two way; c) One channel is directly combined and output by speaker A; the other channel is combined and output by speaker B after passing through their own independent weighting filters. 2. the self-excited elimination method of the high-gain public address system based on space cancellation as claimed in claim 1, is characterized in that, described search optimal weight coefficient

The method is: The first step is to find the phase Let the single-frequency signal from speaker A be

The single-frequency signal from speaker B is in

is the phase caused by space propagation, A ₁ and A ₂ are the amplitudes of the two signals,

is the weighted phase; 256 points are collected in one period, and the sum of the squares of the amplitudes of each point is calculated, which is recorded as emp1, and emp1 can be deduced with

Changes in a sinusoidal law, and the relationship between them can be expressed as

Among them, A, B, and C are three unknown parameters to be determined; four phases are taken

Record the corresponding emp1 ₁ , emp1 ₂ , emp1 ₃ , emp1 ₄ , and get A, B, C by least square method. remember where ε is Gaussian white noise, $Y=\left[\begin{array}{c}{\mathrm{emp}1}_1\\ {\mathrm{emp}1}_2\\ {\mathrm{emp}1}_3\\ {\mathrm{emp}1}_4\end{array}\right],$

$Z=\left[\begin{array}{c}C\\ A\\ B\end{array}\right],$ Then Z=(X ^T X) ^-1 X ^T Y, get A, B, C; and then obtain for:

The second step is to find the amplitude A _i . Let the single-frequency signal from speaker A received by the microphone be

Speaker B is

Take 256 points in one cycle, calculate the sum of the squares of the amplitudes of each point, record it as emp2, keep A ₁ unchanged, change A ₂ , deduce that emp2 changes with A ₂ in a parabolic law, then emp2=aA ₂ ² + bA ₂ +c, where a, b, c are three unknown parameters to be determined. Take four amplitude values A ₂₁ , A ₂₂ , A ₂₃ , A ₂₄ , record the corresponding emp2 ₁ , emp2 ₂ , emp2 ₃ , emp2 ₄ , and get the undetermined parameters a, b, c by least square method. Note emp2＝aA ₂ ² +bA ₂ +c+ε, ε～N(0, σ ² ), where ε is Gaussian white noise, $\mathrm{<span\; class="notranslate">\; Q</span>}<span\; class="notranslate">\; =</span>\left[\begin{array}{c}{\mathrm{<span\; class="notranslate">\; emp</span>}\mathrm{<span\; class="notranslate">\; 2</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}\\ {\mathrm{<span\; class="notranslate">\; emp</span>}\mathrm{<span\; class="notranslate">\; 2</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}\\ {\mathrm{<span\; class="notranslate">\; emp</span>}\mathrm{<span\; class="notranslate">\; 2</span>}}_{\mathrm{<span\; class="notranslate">\; 3</span>}}\\ {\mathrm{<span\; class="notranslate">\; emp</span>}\mathrm{<span\; class="notranslate">\; 2</span>}}_{\mathrm{<span\; class="notranslate">\; 4</span>}}\end{array}\right]<span\; class="notranslate">\; ,</span>$ $\mathrm{<span\; class="notranslate">\; P</span>}<span\; class="notranslate">\; =</span>\left[\begin{array}{ccc}\mathrm{<span\; class="notranslate">\; 1</span>}& {{\mathrm{<span\; class="notranslate">\; A</span>}}_{\mathrm{<span\; class="notranslate">\; twenty\; one</span>}}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}& {\mathrm{<span\; class="notranslate">\; A</span>}}_{\mathrm{<span\; class="notranslate">\; twenty\; one</span>}}\\ \mathrm{<span\; class="notranslate">\; 1</span>}& {{\mathrm{<span\; class="notranslate">\; A</span>}}_{\mathrm{<span\; class="notranslate">\; twenty\; two</span>}}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}& {\mathrm{<span\; class="notranslate">\; A</span>}}_{\mathrm{<span\; class="notranslate">\; twenty\; two</span>}}\\ \mathrm{<span\; class="notranslate">\; 1</span>}& {{\mathrm{<span\; class="notranslate">\; A</span>}}_{\mathrm{<span\; class="notranslate">\; twenty\; three</span>}}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}& {\mathrm{<span\; class="notranslate">\; A</span>}}_{\mathrm{<span\; class="notranslate">\; twenty\; three</span>}}\\ \mathrm{<span\; class="notranslate">\; 1</span>}& {{\mathrm{<span\; class="notranslate">\; A</span>}}_{\mathrm{<span\; class="notranslate">\; twenty\; four</span>}}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}& {\mathrm{<span\; class="notranslate">\; A</span>}}_{\mathrm{<span\; class="notranslate">\; twenty\; four</span>}}\end{array}\right]<span\; class="notranslate">\; ,</span>$ $\mathrm{<span\; class="notranslate">\; K</span>}<span\; class="notranslate">\; =</span>\left[\begin{array}{c}\mathrm{<span\; class="notranslate">\; c</span>}\\ \mathrm{<span\; class="notranslate">\; a</span>}\\ \mathrm{<span\; class="notranslate">\; b</span>}\end{array}\right]<span\; class="notranslate">\; ,</span>$ Then K=(P ^T P) ^-1 P ^T Q can be obtained, and then the amplitude can be obtained as

Images