CN-121985259-A - Local sound amplifying method based on acoustic feedback suppression technology

CN121985259ACN 121985259 ACN121985259 ACN 121985259ACN-121985259-A

Abstract

The invention discloses a local sound amplifying method based on an acoustic feedback suppression technology, and belongs to the field of acoustic feedback suppression of sound amplifying systems. The invention aims to solve the technical problem of poor howling inhibition effect caused by impure reference signals and poor environmental adaptability in the prior art. The method comprises the steps of synchronously acquiring a pure voice reference signal and a mixed audio signal through a bone conduction sensor and an air conduction microphone, measuring an acoustic environment of a sound amplifying space according to a detection signal, adjusting a neural network prediction model, carrying out self-adaptive filtering on the mixed audio signal by adopting the pure voice reference signal to extract a voice component of a speaker, carrying out howling risk analysis according to the adjusted prediction model to generate a potential howling frequency point, carrying out self-adaptive notch processing by utilizing the potential howling frequency point to form a primary sound amplifying signal, and finally applying random micro-delay disturbance to generate a final feedback sound amplifying signal.

Inventors

HU CHENGYUAN
WU ZHENFEI
ZHU ENDE
ZHANG LUJUN

Assignees

深圳波洛斯科技有限公司

Dates

Publication Date: 20260505
Application Date: 20260402

Claims (10)

1. The local sound amplifying method based on the acoustic feedback suppression technology is characterized by comprising the following steps of: S1, synchronous acquisition is carried out on a speaker' S sounding and an ambient sound field by utilizing a bone conduction sensor and an air conduction microphone, and a pure voice reference signal and a mixed audio signal are obtained; S2, carrying out acoustic measurement on the sound amplifying space according to the transmitted detection signal, and adjusting a neural network prediction model according to a measurement result to obtain an environment self-adaptive critical frequency predictor; s3, performing adaptive filtering on the mixed audio signal by adopting the pure voice reference signal to obtain a speaker voice component; s4, carrying out howling risk analysis on the speaker voice component according to the environment self-adaptive critical frequency predictor, and generating potential howling frequency points; S5, performing self-adaptive notch processing on the speaker voice component by utilizing the potential howling frequency point to form a primary suppressed amplified signal; S6, applying random micro-delay disturbance to the amplified signal which is restrained for the first time based on the potential howling frequency point, and generating the amplified signal which is restrained for feedback finally.
2. A local amplification method based on acoustic feedback suppression techniques as claimed in claim 1, the method is characterized in that the step S1 comprises the following steps: S1.1, picking up a pure voice reference signal only containing the sound of a speaker through a bone conduction sensor attached to the skull or mandible of the speaker; S1.2, picking up a mixed audio signal containing the sound of a speaker, the reentry sound of a loudspeaker and the environmental noise through an air-guide microphone arranged at the collar or the head-wearing position of the speaker; s1.3, performing time delay estimation and compensation on the pure voice reference signal and the mixed audio signal, so that the speaker sounding components in the two paths of signals are aligned on a time axis.
3. A local amplification method based on acoustic feedback suppression techniques as claimed in claim 1, the method is characterized in that the step S2 comprises the following steps: S2.1, transmitting a broadband detection signal which covers a sound amplifying target frequency band and is not easy to be perceived by human ears to a sound amplifying space through a loudspeaker array of a sound amplifying system, receiving reflected sound of the detection signal through a gas guide microphone, and calculating reverberation time and frequency response characteristics of the sound amplifying space according to impulse response between the transmitted signal and the received signal; S2.2, taking the calculated reverberation time and frequency response characteristics as input parameters, dynamically adjusting the internal weight and the threshold value of a pre-trained neural network prediction model, so that the neural network prediction model is adapted to the acoustic environment of the current sound amplifying space, and the neural network prediction model is used for outputting the risk probability of howling at each frequency point according to the input audio signal frequency spectrum.
4. A local amplification method based on acoustic feedback suppression technology according to claim 3, wherein the step S2 further comprises dynamically adjusting internal weights and thresholds of the neural network prediction model: Introducing the calculated reverberation time and frequency response characteristics as constraint conditions into a back propagation algorithm of a neural network, so that the model is guided by taking matching of the current acoustic environment as a target in the weight updating process; Under the guidance of the constraint condition, the connection weight and the bias threshold value of each layer of the neural network are subjected to iterative correction through a back propagation algorithm, so that the acoustic features extracted by the feature extraction layer of the model are adaptively matched with the acoustic mode of the current sound amplifying space.
5. A local amplification method based on acoustic feedback suppression techniques as claimed in claim 1, the method is characterized in that the step S3 comprises the following steps: S3.1, taking the pure voice reference signal subjected to delay compensation and spectrum matching as expected input of the adaptive filter, taking the mixed audio signal as original input of the adaptive filter, and initializing filter coefficients; S3.2, iteratively calculating an error signal between a filter output signal and an expected input through an adaptive filtering algorithm, and updating a filter coefficient in real time according to the error signal, so that the filter output gradually approaches a speaker voice component in the mixed audio signal; And S3.3, taking the estimated signal output by the converged adaptive filter as an extracted speaker voice component, and inputting the error signal as a residual reentrant sound component to a feedback inhibition effect evaluation module, wherein the feedback inhibition effect evaluation module judges the current feedback inhibition state according to the energy change trend of the residual reentrant sound component, and when the energy of the residual reentrant sound component continuously rises, the notch depth of the adaptive notch processing in the subsequent step is triggered to be enhanced.
6. A local amplification method based on acoustic feedback suppression techniques as set forth in claim 5, the method is characterized in that the step S3.2 comprises the following steps: Iterative calculation of an error signal between an output signal of the filter and an expected input through an adaptive filtering algorithm, and dynamic adjustment of convergence step length and leakage factor of the adaptive filter according to the amplitude and frequency spectrum characteristics of the error signal; When the bursty pulse interference is detected in the error signal, updating the filter coefficient is temporarily frozen to avoid coefficient divergence, and updating is recovered after the pulse interference is over.
7. A local amplification method based on acoustic feedback suppression techniques as set forth in claim 5, the method is characterized in that the step S3.3 comprises the following steps: Taking the estimation signal output by the converged adaptive filter as an extracted speaker voice component, and inputting the error signal as a residual reentrant sound component to a feedback inhibition effect evaluation module; And the feedback inhibition effect evaluation module judges the current feedback inhibition state according to the energy change trend of the residual reentrant sound component, and triggers the notch depth enhancement of the self-adaptive notch processing in the subsequent step when the energy of the residual reentrant sound component continuously rises.
8. A local amplification method based on acoustic feedback suppression techniques as claimed in claim 1, the method is characterized in that the step S4 comprises the following steps: S4.1, inputting the frequency spectrum data of the speaker voice component to the environment self-adaptive critical frequency predictor, and calculating and outputting howling risk probability values corresponding to all frequency points through forward propagation of the neural network prediction model; S4.2, comparing the howling risk probability value of each frequency point with a plurality of preset risk thresholds, marking the frequency points exceeding the first threshold as potential howling frequency points according to the comparison result, and sequencing the potential howling frequency points according to the risk probability; S4.3, carrying out time sequence analysis on the marked potential howling frequency points, detecting the change trend of the risk probability of each frequency point along with time, marking the frequency point with the continuously increased risk probability as an emergency howling frequency point, and improving the processing priority of the frequency point.
9. A local amplification method based on acoustic feedback suppression techniques as claimed in claim 1, the method is characterized in that the step S5 comprises the following steps: S5.1, configuring corresponding notch depth, notch width and notch quality factors for each potential howling frequency point according to the potential howling frequency point and the corresponding risk level, and distributing larger notch depth and narrower notch width for the frequency point with higher risk level; S5.2, connecting the configured adaptive wave traps in series into a signal processing link according to the frequency point priority order, carrying out notch processing on the voice component of the presenter in sequence, and attenuating only frequency components near the corresponding potential howling frequency point by each wave trap; S5.3, monitoring the actual energy change of each potential howling frequency point in real time in the notch processing process, comprehensively judging by combining the energy change trend of the residual reentrant sound component, and gradually reducing the notch depth of the corresponding notch filter until the notch depth is completely released when the energy of a certain frequency point is reduced below a safety threshold so as to avoid unnecessary damage to sound quality.
10. A local amplification method based on acoustic feedback suppression techniques as claimed in claim 1, the method is characterized in that the step S6 comprises the following steps: S6.1, independently generating a group of pseudo-random time delay sequences for each channel in the multi-channel loudspeaker array according to the distribution range and the risk level of the potential howling frequency points, wherein the maximum time delay amplitude of the pseudo-random time delay sequences is inversely related to the wavelength of the potential howling frequency points covered by the channel, so that the high-frequency sound waves obtain finer time disturbance to effectively destroy the spatial coherence of the high-frequency sound waves; s6.2, copying and distributing the primarily suppressed amplified signals to all channels of the multi-channel loudspeaker array, and applying pseudo-random time delay disturbance of corresponding channels to audio signals of all channels to enable the phase relation of sound waves emitted by the loudspeakers of all channels to be in randomized distribution in space; and S6.3, monitoring the energy change of each potential howling frequency point in the amplified signal subjected to disturbance in real time, and dynamically adjusting the disturbance amplitude and the change rate of the pseudo-random time delay sequence of the corresponding channel when the rising trend of the energy of a certain frequency point is detected.

Description

Local sound amplifying method based on acoustic feedback suppression technology Technical Field The invention discloses a local sound amplifying method based on an acoustic feedback suppression technology, and belongs to the field of acoustic feedback suppression of sound amplifying systems. Background The local sound amplifying system is widely applied to scenes such as education, conferences and stage performances, and has the core functions of picking up the sound of a speaker through a microphone, amplifying the sound and playing the sound through a loudspeaker so as to counteract the attenuation of the sound in space propagation. However, when the sound played back by the speaker is picked up again by the microphone and forms a positive feedback loop, the system will produce howling. Howling not only severely damages the hearing, interferes with normal amplification, but may also cause overload damage to the power amplifier or speaker unit. The conventional acoustic feedback suppression method has respective limitations. The two-stage wave trap needs to detect the howling frequency point and set the notch in advance before the system is used, but the method has long debugging time, can not be quickly adapted to the change of the sound amplifying environment, and can easily generate unexpected howling when the sound field changes. Although the adaptive notch filter can detect and suppress howling in real time, the processing is usually started only after howling has occurred, and acoustic interference at the moment of howling and impact on the system cannot be avoided. Although the method of shift frequency and the like can improve the stability of the system to a certain extent, the gain improving capability of the method is limited, and the requirement of high-fidelity sound amplifying is difficult to meet. Existing adaptive filtering-based feedback suppression techniques typically use the speaker drive signal as a reference in an attempt to subtract the reentrant sound component from the microphone pick-up signal. However, the fundamental disadvantage of this approach is that the reference signal is highly correlated with the desired signal, resulting in difficult filter convergence, large offset, and difficulty in accurately separating the presenter's original speech from re-entry. In addition, the processing parameters of the traditional method are mostly fixed, cannot be dynamically adjusted according to the actual acoustic characteristics of the sound amplifying space, and are unstable in performance under different acoustic environments, and residual feedback still can influence the voice definition. Disclosure of Invention The invention aims to provide a local sound amplifying method based on an acoustic feedback suppression technology, which utilizes a bone conduction sensor and an air conduction microphone to synchronously acquire a pure voice reference signal and a mixed audio signal, carries out acoustic measurement on a sound amplifying space according to a transmitted detection signal, adjusts a neural network prediction model according to a measurement result to obtain an environment self-adaptive critical frequency predictor, carries out self-adaptive filtering on the mixed audio signal by adopting the pure voice reference signal to obtain a speaker voice component, carries out howling risk analysis on the speaker voice component according to the environment self-adaptive critical frequency predictor to generate a potential howling frequency point, carries out self-adaptive notch processing on the speaker voice component by utilizing the potential howling frequency point to form a primary suppressed sound amplifying signal, and applies random micro-delay disturbance on the primary suppressed sound amplifying signal based on the potential howling frequency point to generate a final suppressed feedback sound amplifying signal. According to the invention, through multi-dimensional fusion of bone conduction reference signals, environment self-adaptive prediction, notch processing and micro-delay disturbance, the local sound amplification feedback inhibition with high gain, no loss of sound quality and strong environment adaptability is realized. The aim of the invention can be achieved by the following technical scheme: A local sound amplifying method based on an acoustic feedback suppression technology comprises the following steps: S1, synchronous acquisition is carried out on a speaker' S sounding and an ambient sound field by utilizing a bone conduction sensor and an air conduction microphone, and a pure voice reference signal and a mixed audio signal are obtained; S2, carrying out acoustic measurement on the sound amplifying space according to the transmitted detection signal, and adjusting a neural network prediction model according to a measurement result to obtain an environment self-adaptive critical frequency predictor; s3, performing adaptive filtering on the mixed audio signal b