CN110677796A - Audio signal processing method and hearing aid - Google Patents
Audio signal processing method and hearing aid Download PDFInfo
- Publication number
- CN110677796A CN110677796A CN201910192947.XA CN201910192947A CN110677796A CN 110677796 A CN110677796 A CN 110677796A CN 201910192947 A CN201910192947 A CN 201910192947A CN 110677796 A CN110677796 A CN 110677796A
- Authority
- CN
- China
- Prior art keywords
- audio
- adaptive filter
- howling
- signal
- frequency point
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04R—LOUDSPEAKERS, MICROPHONES, GRAMOPHONE PICK-UPS OR LIKE ACOUSTIC ELECTROMECHANICAL TRANSDUCERS; DEAF-AID SETS; PUBLIC ADDRESS SYSTEMS
- H04R25/00—Deaf-aid sets, i.e. electro-acoustic or electro-mechanical hearing aids; Electric tinnitus maskers providing an auditory perception
- H04R25/30—Monitoring or testing of hearing aids, e.g. functioning, settings, battery power
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04R—LOUDSPEAKERS, MICROPHONES, GRAMOPHONE PICK-UPS OR LIKE ACOUSTIC ELECTROMECHANICAL TRANSDUCERS; DEAF-AID SETS; PUBLIC ADDRESS SYSTEMS
- H04R25/00—Deaf-aid sets, i.e. electro-acoustic or electro-mechanical hearing aids; Electric tinnitus maskers providing an auditory perception
- H04R25/45—Prevention of acoustic reaction, i.e. acoustic oscillatory feedback
- H04R25/453—Prevention of acoustic reaction, i.e. acoustic oscillatory feedback electronically
Landscapes
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Otolaryngology (AREA)
- Neurosurgery (AREA)
- Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Acoustics & Sound (AREA)
- Signal Processing (AREA)
- Circuit For Audible Band Transducer (AREA)
Abstract
The invention discloses an audio signal processing method and a hearing aid, wherein the method comprises the following steps: carrying out howling pre-detection on the forward audio transmission signal; carrying out howling frequency point detection by adopting a three-step method: closing the adaptive filter, starting a wave trap in a forward transmission circuit, and inhibiting the gain of the final howling frequency point; if no howling frequency point is detected, starting the adaptive filter and performing decorrelation processing; obtaining the convergence state step length, the transition state step length and the steady state step length of a tap weight vector of the adaptive filter by using a recursive least square method; the hearing aid comprises a sound pickup unit, a forward gain unit, a detection unit, a wave trap unit, an adaptive filter and an audio output unit. The invention improves the detection precision of the howling frequency point, reduces the influence of the correlation of the actual audio transmission signal on the performance of the self-adaptive filter, and avoids the influence of the fixed step length on the convergence speed and the convergence precision of the filter.
Description
Technical Field
The present invention relates to the field of audio signal processing technology, and in particular, to an efficient audio signal processing method and a hearing aid using the same.
Background
The closed loop exists in an audio transmission path of the hearing aid, positive feedback is generated when the phase of feedback audio is the same as that of external audio picked up by a sound pick-up according to the stable judgment basis of a Nyquist closed loop system, on the basis of the positive feedback, the amplitude of the feedback audio and the amplitude of the external audio picked up by the sound pick-up are both larger than 1, the closed loop system is unstable, and the feedback audio is repeatedly, circularly amplified and accumulated to form self-oscillation to generate howling.
Howling exists in the use process of various acoustic devices, the hearing aid is small in size, low in power consumption and limited in computation amount, and a lot of difficulties exist in the process of designing and inhibiting the howling. Such as personal characteristics of the patient, physical characteristics of the hearing aid, malfunction of the hearing aid or eardrum, and changes in the acoustic environment; furthermore, the wide dynamic compression algorithm of the hearing aid provides higher gain in the low frequency band and reduces the gain in the high frequency band, so that the user can generate howling in a quiet environment or when the input signal is a low-frequency signal. The wide dynamic compression algorithm is a necessary algorithm of the hearing aid, especially in the hearing aid used by the elderly users.
In the prior art, algorithms for solving howling suppression are numerous, and a gain attenuation method, a wave trap method and adaptive filtering are adopted. The gain attenuation method mainly aims at reducing the howling gain in an audio transmission channel, and the gain attenuation method is low in power consumption; the trap filter method is originally aimed at howling caused by static factors, and the inhibition effect on the howling generated by a variable path is not good; both of these algorithms have an effect on the output gain of the hearing aid.
The adaptive filtering algorithm eliminates howling by adjusting the frequency response close to the fed back audio and then subtracting the output of the filter from the output of the microphone. A common estimation algorithm in the adaptive filtering algorithm is that a standard minimum mean square error algorithm is a standard minimum mean square error algorithm, and the standard minimum mean square error method normalizes the update value of the weight vector of each iteration relative to the signal energy on the basis of the minimum mean square error so as to reduce the influence of the amplitude fluctuation of input data on the stability of the algorithm.
In the process that the existing adaptive filtering algorithm approaches to the most stable solution, the convergence performance is poor, and the convergence precision and speed are mainly represented to be insufficient. The existing adaptive filter is designed based on that input signals are mutually independent and the signals are all related in practice, so that the filtering performance of the adaptive filter is obviously reduced based on the independence assumption. Meanwhile, when the howling frequency of the audio signal is detected, the calculation complexity is high, the detection precision is low, and the false alarm rate is high.
Disclosure of Invention
Aiming at the problems of poor convergence performance, poor adaptation to sound environment and low howling frequency detection precision of the existing adaptive filtering algorithm, an audio signal processing method and a hearing aid adopting the method are provided, a three-step method is provided to ensure the detection precision of howling frequency points, if the howling is not detected in a forward circuit, an adaptive filter is started, the adaptive filter effectively reduces the influence of the correlation of actual audio transmission signals on the performance of the adaptive filter by adding decorrelation factors, and meanwhile, a tap weight vector of the adaptive filter adopts a variable-step iterative algorithm, so that the influence of a fixed step on the convergence speed and the convergence precision of the filter is effectively avoided.
In a first aspect, an audio signal processing method for processing an audio signal picked up by a microphone of a hearing aid is provided, comprising the steps of:
s1, carrying out howling pre-detection on the forward audio transmission signal;
s2, performing howling frequency point detection by adopting a three-step method:
sa. selects N larger frequency points in the power energy amplitude in the current frame as candidate howling frequency points,
sb. obtaining the stability parameter S of all candidate howling frequency points by calculating the amplitude power spectrum of the audio signal, if the stability parameter S is larger than the detection threshold STIf so, the candidate howling frequency point is an estimated howling frequency point,
wherein, the stability parameter is calculated by the power spectrum and the influence factor of continuous M frames of signals containing the current frame of signals at different frequency points,
sc., according to the change rate of the peak power slope of the continuous M frames of the estimated howling frequency point, if the change rate of the peak power slope is greater than zero, judging the estimated howling frequency point as the final howling frequency point;
s3, closing the adaptive filter, starting a wave trap in a forward transmission circuit, and restraining the gain of the final howling frequency point;
s4, if the forward audio transmission signal does not detect the howling frequency point, the self-adaptive filter is started, and the audio input signal of the self-adaptive filter is subjected to decorrelation processing;
and S5, obtaining a convergence state step length, a transition state step length and a steady state step length of the tap weight vector of the adaptive filter respectively corresponding to a first stage, a second stage and a third stage by using a recursive least square method, wherein the first step length value is larger than the second step length value, and the third step length value is between the first step length value and the second step length value.
With reference to the first aspect, in a first implementation case, the step Sb includes a step of obtaining the flatness parameter:
sb1, windowing the audio transmission signal in a frame mode, and carrying out Fourier transform on the audio transmission signal subjected to windowing to obtain a signal power spectrum;
sb2, acquiring power spectrums and average values of N candidate howling frequency points including continuous M frames including a current frame;
and Sb3, acquiring a stability parameter S by combining influence factors of the candidate howling frequency points according to the power spectrum and the average value thereof, wherein the influence factors are determined by the number of the frequency points on two sides of the selected candidate howling frequency points.
With reference to the first aspect, in a second implementation case, the step S1 includes the steps of:
and judging whether a blank segment exists in the audio transmission signal of the current frame, if so, discarding the current frame, and continuously detecting the next frame.
With reference to the first aspect, in a second implementation case, the step S4 includes the steps of:
adding a decorrelation factor to the adaptive filter, so that the input signal of the adaptive filter at the time n is:
S(n)＝X(n)-α(n)X(n-1) (1)
equation (1), where s (n) is the input audio signal after decorrelation, x (n) is the actual input audio signal of the adaptive filter, α (n) is the decorrelation factor, and the decorrelation factor α (n) is:
with reference to the second implementation case, in a third implementation case, the step S5 includes the steps of:
s51, updating the tap weight vector of the adaptive filter using the following formula:
in equation (3), w (n) is the tap weight vector of the adaptive filter at time n, y (n) is the hearing aid output signal, e (n) is the forward transmission error signal, and δ is a small positive value.
S52, obtaining the convergence step size of the tap weight vector of the adaptive filter:
in the formula (4), phi1For step size of convergence, #c(n) is the transition state step size, #2For a steady state step size, #c(n) satisfies the following formula
ψc(n)＝ψ1*γ(n) (5)
In the formula (5), γ (n) is a step variable.
In a second aspect, a hearing aid using the audio signal processing method of the first aspect comprises a pick-up unit, a forward gain unit, a detection unit, a wave trap unit, an adaptive filter, and an audio output unit,
the sound pickup unit is used for picking up external sound signals of the hearing aid, converting the voice signals into digital audio signals and transmitting the digital audio signals in the forward circuit;
the forward gain unit is used for amplifying the power of the audio transmission signal in the forward circuit;
the detection unit is used for detecting whether the audio transmission signal in the forward circuit has howling, and in a certain time slice, if the howling exists, the wave trap is started, and if the howling does not exist, the adaptive filter is started;
the adaptive filter is used for decorrelating the input audio transmission signal, tracking and estimating a feedback signal of the output signal of the audio system, specifically executing a variable step size convergence algorithm of normalized least mean square and outputting the feedback signal;
and the audio output unit is used for outputting the processed audio transmission signal and converting the digital audio signal into a sound signal for outputting.
With reference to the second aspect, in a first implementation case, the hearing aid further includes a delay unit in the forward circuit for synchronizing the input audio transmission signal with the feedback signal.
The audio signal processing method and the hearing aid ensure the detection precision of the howling frequency points through a three-step method, firstly determine the frequency points with larger amplitude as candidate howling frequency points, then obtain the stability parameters among the candidate howling frequency point frames through windowing and Fourier transformation of audio transmission signals, judge the estimated howling frequency points through a threshold value, finally determine that the predicted howling frequency points with positive inter-frame amplitude power slope change rate, and then add a wave trap to the howling frequency points to reduce the audio gain. The three-step method effectively reduces the false alarm rate. If the howling frequency point is not detected in the forward circuit, the adaptive filter is started, the adaptive filter effectively reduces the influence of the correlation of the actual audio transmission signal on the performance of the adaptive filter by adding a decorrelation factor, and meanwhile, the tap weight vector of the adaptive filter adopts a variable step size iterative algorithm, so that the influence of a fixed step size on the convergence speed and the convergence precision of the filter is effectively avoided.
Drawings
In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art that other drawings can be obtained according to these drawings without creative efforts.
FIG. 1 is a flow chart of an audio signal processing method according to the present invention;
FIG. 2 is a schematic flow chart of a "three-step method" in an audio signal processing method according to the present invention;
FIG. 3 is a flow diagram illustrating the sub-steps of a "three-step" audio signal processing method according to the present invention;
fig. 4 is a schematic flow chart illustrating the sub-step of step S5 in the audio signal processing method according to the present invention;
FIG. 5 is a schematic diagram of a signal processing model of an audio signal processing method according to the present invention;
fig. 6 is a schematic diagram of the logic components of a hearing aid according to the present invention;
the part names indicated by the numbers in the drawings are as follows: 110-sound pickup unit, 120-forward gain unit, 130-detection unit, 140-wave trap unit, 150-adaptive filter, 160-delay unit, 170-audio output unit.
Detailed Description
The technical solutions in the present invention will be described clearly and completely with reference to the accompanying drawings, and it is obvious that the described embodiments are only some embodiments of the present invention, not all embodiments. Other embodiments, which can be obtained by those skilled in the art based on the embodiments of the present invention without any creative effort, are within the protection scope of the present invention.
In the prior art, a closed loop exists in an audio signal transmission path of a hearing aid, on the basis of positive feedback, the amplitude values of feedback audio and external audio picked up by a sound pick-up are both larger than 1, a closed loop system is unstable, and the feedback audio is repeatedly, circularly amplified and accumulated to form self-oscillation, namely howling. In actual use, the audio transmission path of the audio output unit 170 is generally isolated from the receiving path of the sound pickup unit 110, and the positive feedback is destroyed. This requires a hearing aid with high sealing, but it is difficult to prevent sound leakage from the audio transmission unit, and howling is not avoided.
In theoretical studies of the hearing aid to suppress howling, the trap unit 140 is usually added at the frequency point where howling occurs to reduce the gain of howling. However, when howling is detected, the accuracy is not sufficient, the false alarm rate is high, and therefore, the trap unit 140 is not added at a proper frequency point, the howling gain is reduced, the audio transmission signal is distorted, and the performance of the hearing aid is further affected.
The adaptive filtering algorithm eliminates howling by adjusting the frequency response close to the fed back audio and then subtracting the output of the filter from the output of the microphone. A common estimation algorithm in the adaptive filtering algorithm is a standard minimum mean square error algorithm, and the standard minimum mean square error method normalizes the weight vector update value of each iteration relative to signal energy on the basis of the minimum mean square error so as to reduce the influence of input data amplitude fluctuation on algorithm stability.
Meanwhile, in the process that the existing adaptive filtering algorithm approaches to the most stable solution, the convergence performance is poor, and the convergence precision and speed are mainly represented to be insufficient. The existing adaptive filter 150 is designed based on the fact that the input signals are independent from each other and the signals are correlated in practice, so that the filtering performance of the adaptive filter 150 is obviously reduced based on the assumption of independence.
In view of the above problems, an audio signal processing method and a hearing aid using the same are provided.
Fig. 1 is a schematic flow chart of an audio signal processing method according to the present invention, and fig. 1 is a schematic flow chart of an audio signal processing method, for processing an audio signal picked up by a microphone of a hearing aid, including the steps of:
and S1, carrying out howling pre-detection on the forward audio transmission signal.
Preferably, step S1 includes sub-step S11: and judging whether blank segments exist in the audio transmission signal of the current frame, if so, discarding the current frame, and continuously detecting the next frame.
The sound signal received by the sound pickup unit 110 includes a voice signal, and a blank segment occurs at the place of a voice interval, but a blank segment does not occur in a howling sound frame. By utilizing the characteristic that a part of voice signal frames have blank segments, the voice signal frames with the blank segments are not subjected to howling frequency point detection, so that the complexity of the method is reduced, the loss of important information is reduced, and the reduction of the false alarm rate is facilitated.
The precision of howling frequency point detection cannot be guaranteed based on a single detection characteristic of an audio transmission signal, for example, a voice signal has a resonance peak, the power amplitude of the voice signal is large, and the voice signal often contains important voice information, and if the trap unit 140 is added at the frequency points of the resonance peak, the gain is reduced, and the user loses important information.
S2, performing howling frequency point detection by using a three-step method, as shown in fig. 2, where fig. 2 is a schematic flow diagram of the three-step method in the audio signal processing method provided by the present invention.
Sa., selecting N larger frequency points in the power energy amplitude in the current frame as candidate howling frequency points.
In the frequency domain, the generated howling has a sinusoidal characteristic, and has a high amplitude power at a howling frequency point, and a frequency point with a large amplitude power needs to be selected initially.
Sb. obtaining the stability parameter S of all candidate howling frequency points by calculating the amplitude power spectrum of the audio signal, if the stability parameter S is larger than the detection threshold STAnd the candidate howling frequency point is an estimated howling frequency point, wherein the stability parameter is calculated by the power spectrum and the influence factor of continuous M frames of signals containing the current frame of signals at different frequency points.
Preferably, as shown in fig. 3, fig. 3 is a schematic flow chart of sub-steps of a "three-step method" in the audio signal processing method according to the present invention, and the method for obtaining the stationarity parameter S includes the following steps:
sb1, performs frame windowing on the audio transmission signal, and performs fourier transform on the windowed audio transmission signal to obtain a signal power spectrum.
And Sb2, acquiring power spectrums and average values of N candidate howling frequency points including continuous M frames including the current frame.
And Sb3, acquiring a stability parameter S by combining the influence factors of the candidate howling frequency points according to the power spectrum and the average value thereof, wherein the influence factors are determined by the number of the frequency points at two sides of the selected candidate howling frequency points, the stability parameter S is inversely related to the influence factors, the more the number of the selected frequency points at two sides is, the wider the selected frequency range is, and the farther the outermost frequency point is from the candidate howling frequency points at the moment, the smaller the influence of the outermost frequency point on the stability parameter is.
When howling occurs, the stability parameter of the audio signal at the howling frequency point is sharply increased along with the increase of the howling energy, and the stability of the audio signal without the howling is in the characteristic of oscillation change due to the characteristic of carrying different voice information.
Based on the principle, the stability parameter S of the candidate howling frequency point is calculated, and the detection judgment threshold S is setTIf the stability parameter S of the candidate howling frequency point>STIf yes, the candidate howling frequency point is judged to be the estimated howling frequency point, otherwise, the candidate howling frequency point is still the candidate howling frequency point.
Sc., according to the peak power slope change rate of the estimated howling frequency point continuous M frames, if the peak power slope change rate is greater than zero, the candidate howling frequency point is determined to be the final howling frequency point.
Due to the existence of the formants of the voice signals, the peak value of the frequency point is large, and the frequency point is very easy to become a candidate howling frequency point in preliminary screening. The stability parameter S reflects the trend of the amplitude power changing within a certain frame number range in the time domain, and the detected estimated howling frequency point cannot completely eliminate the formant frequency point of the voice signal. The characteristic that the smoothness of an audio signal is in oscillation change due to the characteristic that the audio signal carries different voice information is utilized, the change rate of the slope of the peak power also oscillates along with the change rate, the howling component is a single-frequency narrow-band signal with the energy increasing continuously, and the peak power of the audio signal is in a rising state all the time before the howling is not eliminated in the transmission process.
And S3, turning off the adaptive filter 150, and enabling the wave trap unit 140 in the forward transmission circuit to suppress the gain of the howling frequency point.
And S4, if the howling frequency point is not detected in the forward audio transmission signal, starting the adaptive filter 150, and performing decorrelation processing on the audio input signal of the adaptive filter 150.
Referring to fig. 5, fig. 5 is a schematic diagram of a signal processing model of an audio signal processing method according to the present invention, and a system error signal e (n) is:
in the formula (6), e (n) is the subtraction of the estimated feedback signal
The residual signal after that, ω (n) is the entire audio signal picked up by the pickup, where ω (n) is:
ω(n)＝X(n)+h(n) (7)
in equation (7), h (n) is the actual feedback signal in the audio transmission path, and x (n) is the actual input audio signal of the adaptive filter 150.
Preferably, the decorrelation processing method is as follows:
when the adaptive filter 150 adds a decorrelation factor, the input signal of the adaptive filter 150 at time n is:
S(n)＝X(n)-α(n)X(n-1) (1)
equation (1), where s (n) is the input audio signal after decorrelation, x (n) is the actual input audio signal of the adaptive filter 150, α (n) is the decorrelation factor, and the decorrelation factor α (n) is:
s5, obtaining a convergence state step size, a transition state step size, and a steady state step size corresponding to the tap weight vector of the adaptive filter 150 in the first stage, the second stage, and the third stage by using a recursive least square method, wherein the first step size is larger than the second step size, and the third step size is between the first step size and the second step size.
Preferably, as shown in fig. 4, fig. 4 is a schematic flowchart of the sub-step of step S5 in the audio signal processing method according to the present invention, and step S5 includes the sub-steps of:
s51, updating the tap weight vector of the adaptive filter 150 using the following formula:
in equation (3), w (n) is the tap weight vector of the adaptive filter 150 at time n, y (n) is the hearing aid output signal, e (n) is the forward transmission error signal, and δ is a small positive value.
S52, obtaining the convergence step size of the tap weight vector of the adaptive filter 150:
in the formula (4), phi1For step size of convergence, #c(n) is the transition state step size, #2For a steady state step size, #c(n) satisfies the following formula
ψc(n)＝ψ1*γ(n) (5)
In the formula (5), γ (n) is a step variable.
In a second aspect, a hearing aid, as shown in fig. 6, fig. 6 is a schematic diagram of logic components of a hearing aid according to the present invention, and the method for processing an audio signal according to the first aspect includes a sound pickup unit 110, a forward gain unit 120, a detection unit 130, a wave trap unit 140, an adaptive filter 150, and an audio output unit 170.
And the sound pickup unit 110 is used for picking up external sound signals of the hearing aid, converting voice signals into digital audio signals and transmitting the digital audio signals in a forward circuit.
A forward gain unit 120, configured to amplify the power of the audio transmission signal in the forward circuit.
A detecting unit 130, configured to detect whether there is howling in the audio transmission signal in the forward circuit, within a certain time slice, if there is howling, the trap unit 140 is enabled, and if there is no howling, the adaptive filter 150 is enabled.
The detection unit 130 first performs howling pre-detection on the forward audio transmission signal. And judging whether a blank segment exists in the audio transmission signal of the current frame, if so, discarding the current frame, and continuously detecting the next frame.
The sound signal received by the sound pickup unit 110 includes a voice signal, and a blank segment occurs at the place of a voice interval, but a blank segment does not occur in a howling sound frame. By utilizing the characteristic that a part of voice signal frames have blank segments, the voice signal frames with the blank segments are not subjected to howling frequency point detection, so that the complexity of the method is reduced, the loss of important information is reduced, and the reduction of the false alarm rate is facilitated.
Furthermore, the detecting unit 130 performs howling frequency point detection by using a three-step method:
a. and selecting N larger frequency points in the power energy amplitude value in the current frame as candidate howling frequency points.
In the frequency domain, the generated howling has a sinusoidal characteristic, and has a high amplitude power at a howling frequency point, and a frequency point with a large amplitude power needs to be selected initially.
b. Obtaining the stability parameter S of all candidate howling frequency points by calculating the amplitude power spectrum of the audio signal, and if the stability parameter S is greater than a detection threshold STAnd the candidate howling frequency point is an estimated howling frequency point, wherein the smoothness parameter is calculated by power spectrums and influence factors of continuous M frames of signals containing the current frame of signals at different frequency points.
When howling occurs, the stability parameter of the audio signal at the howling frequency point is sharply increased along with the increase of the howling energy, and the stability of the audio signal without the howling is in the characteristic of oscillation change due to the characteristic of carrying different voice information.
Based on the principle, the stability parameter S of the candidate howling frequency point is calculated, and the detection judgment threshold S is setTIf the stability parameter S of the candidate howling frequency point>STIf yes, the candidate howling frequency point is judged to be the estimated howling frequency point, otherwise, the candidate howling frequency point is still the candidate howling frequency point.
c. And judging the candidate howling frequency point as a final howling frequency point if the change rate of the peak power slope is greater than zero according to the estimated change rate of the peak power slope of continuous M frames of the howling frequency point.
Due to the existence of the formants of the voice signals, the peak value of the frequency point is large, and the frequency point is very easy to become a candidate howling frequency point in preliminary screening. The stability parameter S reflects the trend of the amplitude power changing within a certain frame number range in the time domain, and the detected estimated howling frequency point cannot completely eliminate the formant frequency point of the voice signal. The characteristic that the smoothness of an audio signal is in oscillation change due to the characteristic that the audio signal carries different voice information is utilized, the change rate of the slope of the peak power also oscillates along with the change rate, the howling component is a single-frequency narrow-band signal with the energy increasing continuously, and the peak power of the audio signal is in a rising state all the time before the howling is not eliminated in the transmission process.
And a trap unit 140 for reducing the forward gain of the howling frequency point.
The adaptive filter 150 decorrelates the input audio transmission signal, tracks and estimates a feedback signal of the output signal of the audio system, and specifically executes a variable step size convergence algorithm of normalized least mean square to output the feedback signal.
If the howling frequency point is not detected in the forward audio transmission signal, the adaptive filter 150 is turned on, and decorrelation processing is performed on the audio input signal of the adaptive filter 150.
The decorrelation factor is added to the adaptive filter 150 to determine the input signal of the adaptive filter 150 at time n. The tap weight vector of the adaptive filter 150 is obtained using the recursive least square method, and the tap weight vector of the adaptive filter 150 is updated using equation (3). The convergence step of the tap weight vector of the adaptive filter 150 is obtained by using equation (4).
And an audio output unit 170 for outputting the processed audio transmission signal, converting the digital audio signal into a sound signal, and outputting the sound signal.
Preferably, the hearing aid further comprises a delay unit 160 in the forward circuit for synchronizing the input audio transmission signal with the feedback signal.
The audio signal processing method and the hearing aid provided by the invention ensure the detection precision of the howling frequency points through a three-step method, firstly determine the frequency points with larger amplitude as candidate howling frequency points, then obtain the stability parameters among the candidate howling frequency point frames through windowing and Fourier transformation of audio transmission signals, judge the estimated howling frequency points through a threshold value, finally determine that the frequency points with positive inter-frame amplitude power slope change rate of the estimated howling frequency points are the howling frequency points, and then reduce the audio gain in the howling frequency point trap unit 140, thereby effectively reducing the false alarm rate through the three-step method. If no howling is detected in the forward circuit, the adaptive filter 150 is started, the adaptive filter 150 effectively reduces the influence of the correlation of the actual audio transmission signal on the performance of the adaptive filter 150 by adding a decorrelation factor, and meanwhile, the tap weight vector of the adaptive filter 150 adopts a variable step size iterative algorithm, so that the influence of a fixed step size on the convergence speed and the convergence precision of the filter is effectively avoided.
The present invention is not limited to the above preferred embodiments, and any modifications, equivalent replacements, improvements, etc. within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (7)
1. A method for processing audio signals picked up by a hearing aid microphone, comprising the steps of:
s1, carrying out howling pre-detection on the forward audio transmission signal;
s2, performing howling frequency point detection by adopting a three-step method:
sa. selects N larger frequency points in the power energy amplitude in the current frame as candidate howling frequency points,
sb. obtaining the stability parameter S of all candidate howling frequency points by calculating the amplitude power spectrum of the audio signal, if the stability parameter S is larger than the detection threshold STIf so, the candidate howling frequency point is an estimated howling frequency point,
the stability parameter S is calculated by power spectrums and influence factors of continuous M frames of signals containing current frame signals at different frequency points;
sc., according to the change rate of the slope of the peak power of the continuous M frames of the estimated howling frequency point, if the change rate of the slope of the peak power is greater than zero, judging the estimated howling frequency point as the final howling frequency point;
s3, closing the adaptive filter, starting a wave trap in a forward transmission circuit, and restraining the gain of the final howling frequency point;
s4, if the howling frequency point is not detected by the forward audio transmission signal, the self-adaptive filter is started, and the audio input signal of the self-adaptive filter is subjected to decorrelation processing;
and S5, obtaining a convergence state step length, a transition state step length and a steady state step length of the tap weight vector of the adaptive filter respectively corresponding to a first stage, a second stage and a third stage by using a recursive least square method, wherein the first step length value is larger than the second step length value, and the third step length value is between the first step length value and the second step length value.
2. The audio signal processing method according to claim 1, wherein the step Sb comprises the step of obtaining the stationarity parameter:
sb1, windowing the audio transmission signal in a frame mode, and carrying out Fourier transform on the audio transmission signal subjected to windowing to obtain a signal power spectrum;
sb2, acquiring power spectrums and average values of N candidate howling frequency points including continuous M frames including a current frame;
and Sb3, acquiring a stability parameter S by combining influence factors of the candidate howling frequency points according to the power spectrum and the average value thereof, wherein the influence factors are determined by the number of the frequency points on two sides of the selected candidate howling frequency points.
3. The audio signal processing method according to claim 1, wherein the step S1 includes the steps of:
and judging whether a blank segment exists in the audio transmission signal of the current frame, if so, discarding the current frame, and continuously detecting the next frame.
4. The audio signal processing method according to claim 1, wherein the step S4 includes the steps of:
adding a decorrelation factor to the adaptive filter, so that the input signal of the adaptive filter at the time n is:
S(n)＝X(n)-α(n)X(n-1) (1)
equation (1), where s (n) is the input audio signal after decorrelation, x (n) is the actual input audio signal of the adaptive filter, α (n) is the decorrelation factor, and the decorrelation factor α (n) is:
5. the audio signal processing method according to claim 1, wherein the step S5 includes the steps of:
s51, updating the tap weight vector of the adaptive filter using the following formula:
in formula (3), w (n) is the tap weight vector of the adaptive filter at time n, y (n) is the hearing aid output signal, e (n) is the forward transmission error signal, and δ is a small positive value;
s52, obtaining the convergence step size of the tap weight vector of the adaptive filter:
in the formula (4), phi1For step size of convergence, #c(n) is the transition state step size, #2For a steady state step size, #c(n) satisfies the following formula:
ψc(n)＝ψ1*γ(n) (5)
in the formula (5), γ (n) is a step variable.
6. A hearing aid using the audio signal processing method proposed in the first aspect, comprising a sound pickup unit, a forward gain unit, a detection unit, a wave trap unit, an adaptive filter, and an audio output unit;
the sound pickup unit is used for picking up external sound signals of the hearing aid, converting the voice signals into digital audio signals and transmitting the digital audio signals in the forward circuit;
the forward gain unit is used for amplifying the power of the audio transmission signal in the forward circuit;
the detection unit is used for detecting whether the audio transmission signal in the forward circuit has howling, and in a certain time slice, if the howling exists, the wave trap is started, and if the howling does not exist, the adaptive filter is started;
the adaptive filter is used for decorrelating the input audio transmission signal, tracking and estimating a feedback signal of the output signal of the audio system, specifically executing a variable step size convergence algorithm of normalized least mean square and outputting the feedback signal;
and the audio output unit is used for outputting the processed audio transmission signal and converting the digital audio signal into a sound signal for outputting.
7. The hearing aid according to claim 6, further comprising a delay unit in the forward circuit of the hearing aid for synchronizing the input audio transmission signal with the feedback signal.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910192947.XA CN110677796B (en) | 2019-03-14 | 2019-03-14 | Audio signal processing method and hearing aid |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910192947.XA CN110677796B (en) | 2019-03-14 | 2019-03-14 | Audio signal processing method and hearing aid |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110677796A true CN110677796A (en) | 2020-01-10 |
| CN110677796B CN110677796B (en) | 2021-12-17 |
Family
ID=69068531
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910192947.XA Active CN110677796B (en) | 2019-03-14 | 2019-03-14 | Audio signal processing method and hearing aid |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110677796B (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112003588A (en) * | 2020-08-27 | 2020-11-27 | 重庆邮电大学 | Adaptive signal filtering method based on polymorphic variable step size normalized mean square |
| CN112039498A (en) * | 2020-08-27 | 2020-12-04 | 重庆邮电大学 | Adaptive signal processing method and medium based on polymorphic variable step length least mean square |
| CN112116914A (en) * | 2020-08-03 | 2020-12-22 | 四川大学 | Sound processing method and system based on variable step length LMS algorithm |
| CN113473304A (en) * | 2021-08-17 | 2021-10-01 | 北京小米移动软件有限公司 | Howling suppression method, howling suppression device, earphone and storage medium |
| WO2023024361A1 (en) * | 2021-08-25 | 2023-03-02 | 歌尔科技有限公司 | Howling suppression method and apparatus, and in-ear earphones and storage medium |
| CN115835092A (en) * | 2023-02-15 | 2023-03-21 | 南昌航天广信科技有限责任公司 | Audio amplification feedback suppression method, system, computer and storage medium |
Citations (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102111707A (en) * | 2009-12-29 | 2011-06-29 | Gn瑞声达公司 | A method for the detection of whistling in an audio system and a hearing aid executing the method |
| CN102316395A (en) * | 2010-07-09 | 2012-01-11 | 廖明忠 | Method and device for judging and eliminating howlround |
| CN102447992A (en) * | 2010-10-06 | 2012-05-09 | 奥迪康有限公司 | Method of determining parameters in an adaptive audio processing algorithm and an audio processing system |
| CN102625213A (en) * | 2012-04-05 | 2012-08-01 | 中国科学院声学研究所 | Audio system squeaking processing method and audio system |
| CN103414504A (en) * | 2013-07-09 | 2013-11-27 | 西南交通大学 | Adaptive beamforming method based on convex combination |
| US20150104039A1 (en) * | 2013-10-15 | 2015-04-16 | Electronics And Telecommunications Research Institute | Apparatus and method of suppressing howling |
| EP2680608B1 (en) * | 2011-08-10 | 2016-02-03 | Goertek Inc. | Communication headset speech enhancement method and device, and noise reduction communication headset |
| JP2017067910A (en) * | 2015-09-29 | 2017-04-06 | ブラザー工業株式会社 | Karaoke device |
| CN107257528A (en) * | 2017-06-14 | 2017-10-17 | 山东浪潮云服务信息科技有限公司 | A kind of detection method of uttering long and high-pitched sounds based on weighted spectral entropy |
| CN107291663A (en) * | 2017-06-12 | 2017-10-24 | 华侨大学 | The variable step suppressed applied to acoustic feedback normalizes sub-band adaptive filtering method |
| CN107645696A (en) * | 2016-07-20 | 2018-01-30 | 腾讯科技（深圳）有限公司 | One kind is uttered long and high-pitched sounds detection method and device |
| CN107708048A (en) * | 2017-09-05 | 2018-02-16 | 腾讯科技（深圳）有限公司 | Detection method of uttering long and high-pitched sounds and device, storage medium and electronic installation |
| EP3367378A1 (en) * | 2017-02-23 | 2018-08-29 | 2236008 Ontario, Inc. | Active noise control using variable step-size adaptation |
-
2019
- 2019-03-14 CN CN201910192947.XA patent/CN110677796B/en active Active
Patent Citations (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102111707A (en) * | 2009-12-29 | 2011-06-29 | Gn瑞声达公司 | A method for the detection of whistling in an audio system and a hearing aid executing the method |
| CN102316395A (en) * | 2010-07-09 | 2012-01-11 | 廖明忠 | Method and device for judging and eliminating howlround |
| CN102447992A (en) * | 2010-10-06 | 2012-05-09 | 奥迪康有限公司 | Method of determining parameters in an adaptive audio processing algorithm and an audio processing system |
| EP2680608B1 (en) * | 2011-08-10 | 2016-02-03 | Goertek Inc. | Communication headset speech enhancement method and device, and noise reduction communication headset |
| CN102625213A (en) * | 2012-04-05 | 2012-08-01 | 中国科学院声学研究所 | Audio system squeaking processing method and audio system |
| CN103414504A (en) * | 2013-07-09 | 2013-11-27 | 西南交通大学 | Adaptive beamforming method based on convex combination |
| US20150104039A1 (en) * | 2013-10-15 | 2015-04-16 | Electronics And Telecommunications Research Institute | Apparatus and method of suppressing howling |
| JP2017067910A (en) * | 2015-09-29 | 2017-04-06 | ブラザー工業株式会社 | Karaoke device |
| CN107645696A (en) * | 2016-07-20 | 2018-01-30 | 腾讯科技（深圳）有限公司 | One kind is uttered long and high-pitched sounds detection method and device |
| EP3367378A1 (en) * | 2017-02-23 | 2018-08-29 | 2236008 Ontario, Inc. | Active noise control using variable step-size adaptation |
| CN107291663A (en) * | 2017-06-12 | 2017-10-24 | 华侨大学 | The variable step suppressed applied to acoustic feedback normalizes sub-band adaptive filtering method |
| CN107257528A (en) * | 2017-06-14 | 2017-10-17 | 山东浪潮云服务信息科技有限公司 | A kind of detection method of uttering long and high-pitched sounds based on weighted spectral entropy |
| CN107708048A (en) * | 2017-09-05 | 2018-02-16 | 腾讯科技（深圳）有限公司 | Detection method of uttering long and high-pitched sounds and device, storage medium and electronic installation |
Non-Patent Citations (2)
| Title |
|---|
| A. PANDEY AND V. J. MATHEWS: "Howling Suppression in Hearing Aids Using Least-Squares Estimation and Perceptually Motivated Gain Control", 《2006 IEEE INTERNATIONAL CONFERENCE ON ACOUSTICS SPEECH AND SIGNAL PROCESSING PROCEEDINGS》 * |
| 郭业才，顾弘毅，王超，王婷: "基于FIR滤波器的改进恒定束宽波束形成算法", 《四川大学学报（自然科学版）》 * |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112116914A (en) * | 2020-08-03 | 2020-12-22 | 四川大学 | Sound processing method and system based on variable step length LMS algorithm |
| CN112116914B (en) * | 2020-08-03 | 2022-11-25 | 四川大学 | Sound processing method and system based on variable step length LMS algorithm |
| CN112003588A (en) * | 2020-08-27 | 2020-11-27 | 重庆邮电大学 | Adaptive signal filtering method based on polymorphic variable step size normalized mean square |
| CN112039498A (en) * | 2020-08-27 | 2020-12-04 | 重庆邮电大学 | Adaptive signal processing method and medium based on polymorphic variable step length least mean square |
| CN112039498B (en) * | 2020-08-27 | 2023-11-14 | 重庆邮电大学 | Self-adaptive signal processing method and medium based on polymorphic variable step-length least mean square |
| CN112003588B (en) * | 2020-08-27 | 2023-11-17 | 重庆邮电大学 | Adaptive signal filtering method based on polymorphic variable step length normalized mean square |
| CN113473304A (en) * | 2021-08-17 | 2021-10-01 | 北京小米移动软件有限公司 | Howling suppression method, howling suppression device, earphone and storage medium |
| CN113473304B (en) * | 2021-08-17 | 2024-01-23 | 北京小米移动软件有限公司 | Howling suppression method, device, earphone and storage medium |
| WO2023024361A1 (en) * | 2021-08-25 | 2023-03-02 | 歌尔科技有限公司 | Howling suppression method and apparatus, and in-ear earphones and storage medium |
| CN115835092A (en) * | 2023-02-15 | 2023-03-21 | 南昌航天广信科技有限责任公司 | Audio amplification feedback suppression method, system, computer and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110677796B (en) | 2021-12-17 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110677796B (en) | Audio signal processing method and hearing aid | |
| US7760888B2 (en) | Howling suppression device, program, integrated circuit, and howling suppression method | |
| EP3703052B1 (en) | Echo cancellation method and apparatus based on time delay estimation | |
| US20230298610A1 (en) | Noise suppression method and apparatus for quickly calculating speech presence probability, and storage medium and terminal | |
| US7092529B2 (en) | Adaptive control system for noise cancellation | |
| US8509451B2 (en) | Noise suppressing device, noise suppressing controller, noise suppressing method and recording medium | |
| US20160372102A1 (en) | Howling Suppression Method and Device Applied to an ANR Earphone | |
| JP5927558B2 (en) | Howling detection device, howling suppression device, and howling detection method | |
| CN104050971A (en) | Acoustic echo mitigating apparatus and method, audio processing apparatus, and voice communication terminal | |
| WO2014054314A1 (en) | Audio signal processing device, method, and program | |
| JP5838861B2 (en) | Audio signal processing apparatus, method and program | |
| CN111418004B (en) | Techniques for howling detection | |
| CN103475980A (en) | Self-adaptive acoustic-feedback-eliminating method | |
| US8108210B2 (en) | Apparatus and method to eliminate noise from an audio signal in a portable recorder by manipulating frequency bands | |
| US9628923B2 (en) | Feedback suppression | |
| Gil-Cacho et al. | Wiener variable step size and gradient spectral variance smoothing for double-talk-robust acoustic echo cancellation and acoustic feedback cancellation | |
| US9330677B2 (en) | Method and apparatus for generating a noise reduced audio signal using a microphone array | |
| WO2015078501A1 (en) | Method of operating a hearing aid system and a hearing aid system | |
| GB2512155A (en) | Acoustic echo cancellation | |
| JP3973929B2 (en) | Howling detection device | |
| CN115278465A (en) | Howling suppression method and device, sound box and sound amplification system | |
| WO2009070850A1 (en) | Entrainment resistant feedback cancellation | |
| CN115175063A (en) | Howling suppression method and device, sound box and sound amplification system | |
| JP2011254420A (en) | Echo elimination method, echo elimination device, and echo elimination program | |
| Karthik et al. | Speech enhancement using gradient based variable step size adaptive filtering techniques |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| CB02 | Change of applicant information |
Address after: 518115 No. 25 Industrial Zone, Fenghuang Road, West hang, Henggang, Shenzhen, Longgang, Guangdong Applicant after: Shenzhen Pangao Medical Electronics Co.,Ltd. Address before: 518115 No. 25 Industrial Zone, Fenghuang Road, West hang, Henggang, Shenzhen, Longgang, Guangdong Applicant before: SHENZHEN PANGAO ELECTRONIC Co.,Ltd. |
|
| CB02 | Change of applicant information | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |