JP2006262339A - Loudspeaker apparatus - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a loudspeaker apparatus capable of quickly and effectively suppressing howling while maintaining the sound quality of sound as a loudspeaking object. <P>SOLUTION: The loudspeaker apparatus 1 is provided with a loudspeaking section 5 wherein a speaker 3 and a microphone 2 are connected and an input sound signal picked up by the microphone 2 is loudly spoken and fed to the speaker 3; a simulation signal generator 7 for generating a simulation signal for simulating a feedback sound fed back from the speaker 3 to the microphone 2; and a subtractor 74 for generating a residual signal resulting from subtracting the simulation signal generated by the simulation signal generator 7 from the input sound signal from the microphone 2, and providing an output to the loudspeaking section 5. Further, the apparatus includes a howling frequency detector 6 that detects a frequency of a howling sound component included in the input sound signal from the microphone 2, and a band limit 71 for placing a limit on frequencies of the simulation signal for simulating the feedback sound only at a frequency band including the frequency detected by the howling frequency detector 6 and in the vicinity of the frequency. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

  The present invention relates to a loudspeaker equipped with a howling canceller for preventing howling caused by a feedback sound being input via a feedback transmission system from a speaker to a microphone.

  A loudspeaker amplifies an input audio signal input from a microphone by a loudspeaker circuit and inputs the amplified signal to a speaker. In this loudspeaker, a closed loop is formed in a path from the speaker to the microphone and is output from the speaker. It is well known that howling occurs when the feedback sound signal input to the microphone is repeatedly amplified.

  In order to prevent such howling, it has been proposed to use a howling canceller using a notch filter for a loudspeaker. In such a loudspeaker, when a howling frequency detector detects a frequency at which howling occurs in an input audio signal from a microphone, a notch filter is set so as to attenuate a sound component in the vicinity of the frequency band including the frequency. The As a result, the howling component is removed from the input audio signal from the microphone by the notch filter and input to the amplifier circuit unit (see, for example, Patent Document 1).

As another method, it has been proposed to use a howling canceller for a loudspeaker that generates a simulated signal simulated by a feedback sound signal using an adaptive filter and removes the simulated signal from an input signal from a microphone (non-patent document). Reference 1). In this howling canceller, the same audio signal as the audio signal input to the speaker is input to the delay unit, and this audio signal is delayed by the delay unit by a delay time corresponding to the delay time from the speaker to the microphone. A simulated signal is generated by performing a convolution operation on the delayed signal with a filter coefficient. Then, a residual signal obtained by removing the simulated signal from the input voice signal from the microphone by the subtraction unit is output to the loudspeaker circuit unit. Then, the residual signal amplified by the loudspeaker circuit unit is input to the speaker and emitted from the speaker. Here, a residual signal is input to the adaptive filter as a reference signal, and a filter coefficient (filter characteristic) is used by using a known adaptive algorithm (for example, LMS (Least Mean Square) algorithm) so that the residual signal is minimized. ) Is corrected. The loudspeaker amplifies the input sound input from the microphone and outputs it from the speaker. In this loudspeaker, a closed loop is formed in the path from the speaker to the microphone, and the sound is output from the speaker and input to the microphone. It is a well-known problem that feedback occurs when the feedback sound signal is amplified many times.
To solve this problem, an adaptive filter is used to obtain a simulated transfer function that simulates the transfer function from the microphone to the speaker, and a simulated signal that simulates the feedback sound signal is calculated using the simulated transfer function. It is known to use a howling canceller that removes a pseudo signal from the input signal of a loudspeaker. In this howling canceller, a delay time from the speaker to the microphone is added to the amplified input signal by the delay unit. A pseudo signal is calculated by the adaptive filter by performing a convolution operation using the input signal to which the delay time is given and the adaptive filter coefficient as a simulated transfer function. Then, an error signal obtained by removing the pseudo signal from the input signal via the microphone is output to the amplifying unit, reproduced, and emitted from the speaker. Here, an error signal is input to the adaptive filter as a reference signal. Then, the adaptive filter coefficient is calculated using the correction amount calculated using a known adaptive algorithm (for example, LMS (Least Mean Square) algorithm) and the filter coefficient before update so that the error signal is minimized. Updated. For this reason, the filter coefficient of the adaptive filter is adapted to the transfer function of the feedback transfer system from the speaker to the microphone, and the filter coefficient simulates the transfer function of the feedback transfer system. For this reason, the signal processed by the adaptive filter, that is, the simulation signal, approximates the feedback sound signal. As a result, the feedback sound signal component can be removed from the residual signal, so that howling can be prevented.
Japanese Patent No. 3097376 Inazumi, Imai, Konishi: “Preventing howling in loudspeakers using the LMS algorithm”, Proc. 417-418 (1991, 3)? ? ?

  In a conventional howling canceller using a notch filter, as described above, a process of removing a signal component in a frequency band detected as a howling component is performed on an audio signal input from a microphone. For this reason, there has been a problem that the sound quality of the processed audio signal is deteriorated. That is, for example, when an audio signal component for the purpose of loudening such as a user's voice (speaker's speech) using a microphone is included in a frequency band detected as a howling component, the audio component in this frequency band Is attenuated by the notch filter, the audio signal for the purpose of loudening is removed. For this reason, the sound quality of the sound that is the purpose of the sound amplification is deteriorated.

  Further, in a conventional howling canceller using an adaptive filter, the adaptive filter generates a simulated signal by simulating a transfer function of a feedback transfer system from a speaker to a microphone. In such a howling canceller, howling is removed and prevented by subtracting the simulated signal from the input voice signal from the microphone. For this reason, if there is an estimation error in the transfer function of the feedback transfer system in the adaptive filter, the simulated signal is removed from the audio signal input from the microphone, thereby removing even the audio signal component for the purpose of loudening. There is a case. For this reason, the sound quality of the sound signal for the purpose of sound amplification may be impaired.

  In addition, since adaptation of the adaptive filter takes some time, there is a problem in that it cannot be immediately adapted to howling that occurs rapidly.

  In order to solve the above-described problems, an object of the present invention is to provide a loudspeaker that can quickly and effectively suppress howling while maintaining the sound quality of the voice that is the purpose of the loudspeaker.

  In order to solve the above problems, the present invention employs the following means.

  (1) In the present invention, a speaker and a microphone are connected, and a sound-amplifying unit that amplifies an input audio signal input from the microphone and supplies the sound to the speaker, and a simulation that simulates feedback sound returned from the speaker to the microphone A simulation signal generation unit that generates a signal, and a subtraction unit that generates a residual signal obtained by subtracting the simulation signal generated by the simulation signal generation unit from an input voice signal from the microphone and outputs the residual signal to the loudspeaker unit A loudspeaker comprising: a howling frequency detection unit for detecting a frequency of a howling voice component included in an input voice signal from the microphone; and a frequency detected by the howling frequency detection unit. A band limiting unit that limits the generation of a simulated signal that simulates the feedback sound only in the nearby frequency band. The features.

  According to the above configuration, the frequency of the howling sound component included in the input sound signal from the microphone is detected by the howling frequency detection unit. Then, the simulation signal generating unit generates a simulation signal that simulates the feedback sound that is fed back from the speaker to the microphone. Here, the band limiting unit is limited to generate a simulation signal that simulates the feedback sound only in the frequency band in the vicinity of this frequency including the frequency detected by the howling frequency detection unit. Then, the subtracting unit subtracts the simulated signal from the input voice signal from the microphone to generate a residual signal, and the residual signal is output to the loudspeaker. As described above, since only the simulated signal only in the frequency band in the vicinity of this frequency including the frequency detected by the howling frequency detection unit is subtracted, it is possible to effectively prevent the removal of the audio signal component to be amplified. It is possible to remove the howling component.

  (2) In the loudspeaker according to the present invention, the band limiting unit includes a frequency near the frequency including a frequency detected by the howling frequency detection unit among audio signals corresponding to the audio signal input to the speaker. A band-pass filter that passes only frequency band components, and the simulation signal generation unit generates a simulation signal using the residual signal input from the subtraction unit and the audio signal that has passed through the band-pass filter. An adaptive filter is provided. With this configuration, in the bandpass filter, only the component in the frequency band near this frequency including the frequency detected by the howling frequency detection unit of the audio signal input to the speaker is passed. Then, the adaptive filter generates a simulation signal using the residual signal input from the subtraction unit and the audio signal that has passed through the bandpass filter. That is, in the adaptive filter, a filter coefficient is calculated using the audio signal that has passed through the bandpass filter, the calculated filter coefficient is self-set, and a simulated signal is generated using the filter coefficient. For this reason, the simulation signal generated by the adaptive filter simulates the feedback sound in the frequency band range near this frequency including the frequency detected by the howling frequency detection unit. As described above, the frequency band in the vicinity of this frequency including the frequency detected by the howling frequency detection unit is obtained by a simple process in which the audio signal input to the speaker after passing through the band-pass filter is input to the adaptive filter. It is possible to limit the generation of the simulation signal that simulates the feedback sound only in the range of.

(3) The present invention connects a speaker and a microphone, and amplifies an input voice signal input from the microphone and supplies the speaker to the speaker;
A simulation signal generation unit that generates a simulation signal that simulates a feedback sound that is fed back from the speaker to the microphone, and a residual signal obtained by subtracting the simulation signal generated by the simulation signal generation unit from an input audio signal from the microphone A subtracting unit that generates and outputs to the loudspeaker unit, and a howling frequency detection unit that detects a frequency of a howling voice component included in an input voice signal from the microphone; A band limiting unit that limits the subtracting unit to subtract only a component of a frequency band in the vicinity of the frequency including the frequency detected by the howling frequency detecting unit from the simulated signal. .

  According to the above configuration, the frequency of the howling sound component included in the input sound signal from the microphone is detected by the howling frequency detection unit. Then, the simulation signal generating unit generates a simulation signal that simulates the feedback sound that is fed back from the speaker to the microphone. Then, the subtracting unit subtracts the simulated signal from the input voice signal from the microphone to generate a residual signal, and the residual signal is output to the loudspeaker. Here, the band limiting unit limits the simulation signal so that only the frequency band component in the vicinity of this frequency including the frequency detected by the howling frequency detection unit is subtracted by the subtraction unit. As described above, since only the components in the frequency band in the vicinity of this frequency including the frequency detected by the howling frequency detector are subtracted from the simulated signal, it is possible to effectively remove the audio signal component to be amplified. It is possible to remove the howling component while preventing it.

  (4) The present invention is a loudspeaker device including a loudspeaker unit that connects a speaker and a microphone, and that loudspeaks an input voice signal input from the microphone and supplies the voice to the speaker. The input voice signal from the microphone A howling frequency detection unit for detecting the frequency and phase of the component of the howling speech contained in the signal, and generating a cancellation signal having the same frequency as the frequency detected by the howling frequency detection unit and an opposite phase to the detected phase And a cancellation signal generation unit, and an addition unit that adds the cancellation signal generated by the cancellation signal generation unit to the input audio signal from the microphone.

  According to the above configuration, the frequency and phase of the howling sound component included in the input sound signal from the microphone are detected by the howling frequency detection unit. Then, a cancellation signal having the same frequency as the frequency detected by the howling frequency detection unit and having a phase opposite to the detected phase is generated by the cancellation signal generation unit. The cancellation signal generated by the cancellation signal generation unit is added to the input audio signal from the microphone by the addition unit. By the addition of the cancellation signal, the howling component in the audio signal input from the microphone is canceled, and this howling component can be removed from the input audio signal from the microphone. In addition, a cancellation signal is generated and added to the input audio signal from the microphone by the adding unit, so that a howling component is removed to add the simulated signal generated by processing with the adaptive filter to the input audio signal from the microphone. It is possible to quickly remove the howling voice component as compared with the configuration to be performed.

According to the first aspect of the present invention, the simulated signal that simulates the feedback sound only in the frequency band near the howling frequency including the howling frequency is excluded from the sound signal from the microphone. According to the third aspect of the present invention, only the components in the frequency band in the vicinity of the howling frequency including the howling frequency in the simulated signal are excluded from the sound signal from the microphone. For this reason, by removing the simulation signal from the input voice signal from the microphone, it is possible to prevent the voice signal in the other frequency band from being removed from the frequency band near the howling frequency. As a result, it is possible to effectively remove the audio signal for the purpose of loudening while removing the howling component from the input audio signal from the microphone. Therefore, the howling component is effectively removed from the audio signal input from the microphone. While removing, it is possible to effectively prevent the deterioration of the sound quality of the voice that is the purpose of the sound amplification.
According to the fourth aspect of the present invention, the howling component in the audio signal input from the microphone is canceled by adding the cancellation signal, and this howling component can be removed from the input audio signal from the microphone. . For this reason, while removing the howling component, it is possible to effectively prevent the deterioration of the audio signal component that is the purpose of the sound enhancement by removing the sound signal component that is the purpose of the sound enhancement. In addition, a cancellation signal is generated and added to the input audio signal from the microphone by the adder, so that the howling component is removed, and the simulated signal generated by processing with the adaptive filter is converted into the input audio signal from the microphone. It is possible to provide a loudspeaker that quickly removes a howling voice component as compared to the configuration of adding.

  Embodiments of the present invention will be described below with reference to the drawings.

(First embodiment)
The loudspeaker according to the first embodiment of the present invention is equipped with a howling canceller using an adaptive filter, thereby preventing howling. Specifically, the loudspeaker detects the presence of howling and the frequency (howling frequency) from the audio signal input from the microphone, and operates the adaptive filter only in the frequency band near the howling frequency including this howling frequency. Thus, the transfer function of the feedback transfer system of the feedback sound from the speaker to the microphone is simulated. The simulated signal processed by the adaptive filter is removed from the audio signal input from the microphone. As a result, in the first embodiment, the howling component can be removed from the input audio signal from the microphone while effectively eliminating the audio signal component for the purpose of loudening such as the voice of the user who uses the microphone. It is possible to do.

  The first embodiment will be described below with reference to FIG. FIG. 1 is a block diagram showing a schematic configuration of a loudspeaker 1 according to the first embodiment. The loudspeaker 1 is connected to a microphone 2, an amplifier 3, and a speaker 43, and includes a signal processor 4, a power amplifier 5, a howling frequency detector 6, and a howling canceller 7.

  The microphone 2 inputs a voice to be amplified such as a voice of a speaker who uses the microphone 2 from the outside of the apparatus as a microphone input signal. This signal is input to the loudspeaker 1 via the input terminal 8. A signal input to the loudspeaker 1 is A / D converted by an A / D (Analog / Digital) converter (not shown), input to a howling frequency detection unit 6, and a signal processing unit via a howling canceller 7. 4 is input.

  The speaker 3 changes the analog audio signal input from the loudspeaker 1 through the output terminal 9 into sound and emits the sound. The speaker 3 is disposed in the same acoustic space as the microphone 2. For this reason, the sound emitted from the speaker 3 is transmitted by the feedback transmission system 100 which is an acoustic transmission path from the speaker 3 to the microphone 2 and is input to the microphone 2 again as a feedback sound. For this reason, the microphone input signal includes an audio signal s (k) and a feedback sound signal d (k) for the purpose of loudening.

  The signal processing unit 4 performs a process of adjusting an audio signal such as equalizing on the audio signal input via the input terminal 8 and the howling canceller 7. The signal processing unit 4 outputs the audio signal subjected to this processing to the speaker 3 as an audio signal x (k). The signal output from the signal processing unit 4 is converted into an analog audio signal by a D / A converter (not shown) and then input to the power amplifier 5. The power amplifier 5 corresponds to the loudspeaker of the present invention, amplifies the signal level of the input analog audio signal, and inputs it to the speaker 3 through the output terminal 9.

  The howling frequency detection unit 6 inputs an input audio signal from the microphone 2 and executes a process (howling detection process) of detecting the presence / absence of a howling component and a howling frequency in the input signal at predetermined time intervals. In this howling detection processing, for example, the howling frequency detection unit 6 performs frequency analysis such as FFT (Fast Fourier Translation) processing on the input signal to calculate a frequency spectrum, and then a spectrum (peak Value). Then, howling frequency detection unit 6 calculates the average value of each spectrum for each frequency band, and detects that a howling component is included in the input signal when the peak value exceeds the threshold value (predetermined level) from the average value. To do. The frequency corresponding to the peak value detected by the howling frequency detection unit 6 as this howling component is defined as the howling frequency. A predetermined narrow frequency band including this howling frequency (a frequency band near the howling frequency) is a howling frequency band.

  Since the howling component is repeatedly amplified by the power amplifier 5, the amplitude is increased. Therefore, when the peak value exceeds the threshold value by a larger value than the average value, there is a high possibility that the component corresponding to the peak is the howling component. . For this reason, it is possible to accurately detect the presence / absence of the howling component and the howling frequency band by a simple process of simply determining whether or not the peak value exceeds the threshold value (predetermined level) from the average value.

  When the howling frequency detection unit 6 detects that a howling component is included in the input signal, the howling frequency detection unit 6 notifies a howling frequency band to a howling canceller 7 (a bandpass filter 71 described later). Further, when the howling frequency detection unit 6 does not detect that a howling component is included in the input signal, the howling frequency detection unit 6 notifies the howling canceller 7 of the howling frequency band as “0”.

  The howling canceller 7 includes a band pass filter 71, a delay unit 72, an adaptive filter 73, and a subtraction unit 74. The bandpass filter 71 corresponds to the band limiting unit of the present invention, receives the audio signal x (k) from the signal processing unit 4, and extracts only the audio signal having the set frequency from the audio signal x (k). To the delay unit 72. As described above, when the howling frequency detection unit 6 detects that a howling component is included in the input signal, the bandpass filter 71 is notified of the howling frequency band detected by the howling frequency detection unit 6. On the other hand, if a howling component is not detected in the input signal, the howling frequency band is notified as “0”. The band pass filter 71 self-sets the characteristic so as to pass the signal only in the notified frequency band.

  Therefore, when the howling frequency detection unit 6 detects that a howling component is included in the input signal, the howling frequency detection unit 6 passes the bandpass filter 71 in the audio signal x (k). Only a narrow frequency band component (audio signal x ′ (k)) including the detected howling frequency is obtained. On the other hand, when the howling frequency detector 6 detects that the howling component is not included in the input signal, the band-pass filter 71 does not pass the audio signal x (k) at all, and the howling canceller 7 does not function. .

  The delay unit 72 and the adaptive filter 73 simulate the feedback transmission system 100. That is, the delay unit 72 simulates the delay time τ of the feedback sound of the feedback transfer system 100, and the adaptive filter 73 simulates the transfer function h that is the sound propagation characteristic of the feedback transfer system 100.

  Specifically, the delay unit 72 delays the input audio signal x ′ (k) by a delay time τ that simulates the delay time of the feedback transmission system 100, and this delayed audio signal x (k−τ) ′. Is output to the adaptive filter 73.

  The adaptive filter 73 corresponds to the simulation signal generation unit of the present invention, includes a digital filter (typically, a FIR (Finite Impulse Response) filter), estimates the transfer function h of the feedback transfer system 100, and estimates the transfer function. The filter coefficient (filter characteristic) of this digital filter is calculated and set in accordance with h (as simulated). The estimation of the transfer function h and the calculation of the filter coefficient are performed by using the residual signal e (k), which is a signal input from the subtraction unit 74, as a reference signal, and the audio signal x ′ (k) input from the delay unit 72. -Τ) using an adaptive algorithm. As an adaptive algorithm, for example, a learning identification method, an LMS method, a projection method, an RLS method, or the like can be applied. This filter coefficient is calculated every predetermined time interval (for example, every several seconds), and the mean square value J = E [e (k) 2] of the residual signal e (k) (where E [E [· ] Is an expected value]), it is calculated to minimize J.

  As described above, the filter coefficient is calculated based on the audio signal x ′ (k−τ), and the audio signal x ′ (k−τ) is a howling component from the audio signal x (k) input to the speaker 3. Only the estimated frequency band is extracted. For this reason, the calculated filter coefficient simulates the transfer function h only in the howling frequency band including the frequency detected by the howling frequency detection unit 6. As a result, the simulated signal processed by the adaptive filter 73 simulates the feedback sound only in the howling frequency band, and the bandpass filter 71 is limited to generate the simulated signal only in the howling frequency band. Become.

  The adaptive filter 73 performs a convolution operation on the input audio signal x ′ (k−τ) with a self-set filter coefficient (giving filter characteristics) to generate a simulation signal do (k). As described above, the filter function simulates the transfer function h only in the range of the frequency band estimated as the howling component, and the speech signal x ′ (k−τ) is estimated as the howling component from the speech signal x (k). Only the frequency band is extracted. For this reason, the simulation signal do (k) simulates only the howling component of the feedback sound input from the microphone 2. The adaptive filter 73 outputs the generated simulation signal do (k) to the subtraction unit 74.

  The subtracting unit 74 receives the simulated signal do (k) from the adaptive filter 73 and the microphone input signal from the microphone 2 via the input terminal 8. The subtracting unit 74 generates a residual signal e (k) obtained by removing the simulated signal do (k) from the microphone input signal. The subtraction unit 74 inputs the generated residual signal e (k) to the signal processing unit 4 and inputs it to the adaptive filter 73 as a reference signal. As described above, since the simulated signal do (k) simulates only the howling component of the feedback sound, by removing the simulated signal do (k) from the microphone input signal, the howling component is removed while removing the howling component. It is effectively prevented to remove even the audio signal. If the howling frequency detector 6 does not determine that a howling component is included in the input signal, the subtractor 74 is in a pause state, so that the microphone input signal passes through the subtractor 74 as it is, and the signal processor. 4 is input.

  The operation of the loudspeaker 1 will be described below. When the speaker speaks, an audio signal such as a speaker's voice is input to the microphone 2. The microphone input signal input to the microphone 2 is input to the signal processing unit 4 via the input terminal 8 and the subtraction unit 74. Here, the microphone input signal is also input to the howling frequency detection unit 6. The howling frequency detection unit 6 detects the presence of a howling component and the howling frequency from the microphone input signal by the method described above.

  When it is detected that there is a howling component, a howling frequency band including the detected howling frequency is notified to the band-pass filter 71 by the howling frequency detection unit 6. On the other hand, when it is detected that there is no howling component, the howling frequency band is set to “0”, and the howling frequency detection unit 6 notifies the bandpass filter 71 of it. Here, the bandpass filter 71 that has received the notification sets itself so as to pass only the signal in the notified howling frequency band.

  The microphone input signal input to the signal processing unit 4 is subjected to predetermined signal processing such as equalizing. The audio signal x (k) subjected to this signal processing is D / A converted by a D / A converter (not shown), the signal level is amplified by the power amplifier 5, and then sent to the speaker 3 via the output terminal 9. Entered. The audio signal x (k) is input to the band pass filter 71. In the band pass filter 71, only the audio signal x ′ (k) that is a component of the set frequency band in the audio signal x (k) is passed and input to the delay unit 72. That is, when the howling frequency detection unit 6 detects that there is a howling component, only the howling frequency band component detected by the howling frequency detection unit 6 is input to the delay unit 72.

  On the other hand, when the howling frequency detection unit 6 detects that there is no howling component, the howling frequency band is “0”, and therefore the band-pass filter 71 does not pass the audio signal x (k) at all and delays. The audio signal x (k) is not input to the unit 72 at all. In this case, the howling canceller 7 is in a dormant state.

  When the audio signal x ′ (k) is input to the delay unit 72, it is delayed by a delay time τ corresponding to the delay time of the feedback transmission system 100 and input to the adaptive filter 73 as the audio signal x ′ (k−τ). Is done. The adaptive filter 73 generates a simulated signal do (k) by convolving the audio signal x ′ (k−τ) with a filter coefficient. The generated simulation signal do (k) is input to the subtraction unit 74. Further, the adaptive filter 73 calculates a filter coefficient using an adaptive algorithm using the audio signal x ′ (k−τ) and the residual signal e (k), and corrects the filter coefficient with the calculated filter coefficient. The

  As described above, the microphone input signal is input to the subtracting unit 74, and when the simulated signal do (k) is input to the subtracting unit 74, the subtracting unit 74 removes the simulated signal do (k) from the microphone input signal. . The residual signal e (k) obtained by removing the simulated signal do (k) from the microphone input signal is input to the signal processing unit 4 and also input to the adaptive filter 73 as a reference signal. On the other hand, when the howling frequency detection unit 6 detects that there is no howling component, the howling canceller 7 is in a pause state as described above, and the simulation signal do (k) is not input to the subtraction unit 74. . For this reason, in the subtraction unit 74, the microphone input signal passes through as it is and is input to the signal processing unit 4.

  With the above configuration, in this embodiment, when it is detected by the howling frequency detection unit 6 that there is a howling component, only the component of the howling frequency band in the audio signal x (k) is detected by the bandpass filter 71. The audio signal x ′ (k) is passed. Then, the audio signal x ′ (k) is input to the adaptive filter 73 via the delay unit 72. For this reason, the simulation signal do (k) generated by the adaptive filter 73 simulates only the component estimated as the howling component in the feedback sound. As a result, by removing the simulated signal do (k) from the microphone input signal by the subtracting unit 74, while removing the howling component, the sound signal that is the target of the sound amplification is also removed, and the sound signal that is the target of the sound enhancement. It is effectively prevented that the sound quality is degraded. As a result, howling can be effectively prevented while maintaining the sound quality of the voice that is to be amplified.

(Second Embodiment)
Hereinafter, a loudspeaker 1A according to a second embodiment of the present invention will be described with reference to FIG. In the first embodiment, the simulation signal do (k) is limited to generate a simulation signal that simulates the feedback sound only in the howling frequency band. In the second embodiment, the simulation signal do (k) Of these, only the component in the howling frequency band is limited to be subtracted from the input signal from the microphone 2 by the subtracting unit 74.

  Referring to FIG. 1, in the first embodiment, the bandpass filter 71 is arranged before the delay unit 72 in the signal path of the audio signal x (k). In the second embodiment, the adaptive filter The simulation signal do (k) from 73 to the subtraction unit 74 is arranged on the signal path. As a result, the band-pass filter 71 passes the simulation signal do (k) only in the howling frequency band and inputs it to the subtracting unit 74. For this reason, the simulation signal do (k) of only the howling frequency band is subtracted from the input audio signal from the microphone 2 by the subtracting unit 74. The configuration of the second embodiment has the same effect as that of the first embodiment.

(Third embodiment)
A loudspeaker 1B according to a third embodiment of the present invention will be described below with reference to FIG. FIG. 2 is a block diagram showing a schematic configuration of a loudspeaker 1B according to the third embodiment. In the first and second embodiments, howling is prevented using the adaptive howling canceller 7 (using the adaptive filter 73). On the other hand, in the third embodiment, a howling component (howling component signal) cancellation signal in the microphone input signal is generated, and the generated cancellation signal is added to the microphone input signal, whereby the howling component is input to the microphone. It is the structure removed from a signal. The cancellation signal is a sine wave having a howling frequency, and is a signal whose phase is delayed or advanced by “½ period” or “integer multiple of period + ½ period” from the howling component signal.

  In FIG. 2, the loudspeaker 1 </ b> B is different from the loudspeaker 1 in that a howling frequency detector 6 </ b> B described later is provided instead of the howling frequency detector 6, and a howling canceller 10 described later is provided instead of the howling canceller 7. Since other configurations and the operation of the configurations are the same, description thereof is omitted.

  The howling frequency detection unit 6B detects the presence / absence of a howling component and the howling frequency at predetermined time intervals by executing howling detection processing on the microphone input signal in the same manner as in the first embodiment. Then, the howling frequency detection unit 6B detects the phase of the signal of the howling component when detecting that the howling component is present, and generates information (howling component information) indicating the phase, the detected howling frequency, and the level of the howling signal. To do. The howling frequency detection unit 6B outputs howling component information to the howling canceller 10 (a cancellation signal generation unit 11 described later).

  The howling canceller 10 includes a cancellation signal generation unit 11 and an addition unit 12. The cancellation signal generator 11 receives howling component information from the howling frequency detector 6B. The cancellation signal generation unit 11 generates a howling component signal cancellation signal based on this howling component information. This cancellation signal is a signal indicating a sine wave having the same frequency and the same signal level as the howling frequency indicated by the howling component information and a phase delayed by, for example, ½ period from the phase indicated by the howling component information. The cancellation signal generation unit 11 inputs the cancellation signal to the addition unit 12 in synchronization with the howling generation signal component corresponding to the generated cancellation signal being input to the addition unit 12.

  The adder 12 receives the input audio signal from the microphone 2 and the cancellation signal from the cancellation signal generator 11. The adding unit 12 adds a cancellation signal to the input audio signal from the microphone 2. As a result, the howling component signal included in the input audio signal from the microphone 2 is canceled by the cancellation signal and removed from the input audio signal from the microphone 2.

  With the above configuration, when the howling frequency detection unit 6B detects that a howling component is included in the microphone input signal, a cancellation signal is generated and added to the input audio signal from the microphone 2, whereby the howling component signal Is canceled by the cancellation signal. As a result, it is possible to suitably remove the howling component from the microphone input signal while preventing the deterioration of the speech that is the purpose of the speech enhancement by removing even the speech component that is the purpose of the speech enhancement. In the third embodiment, a cancellation signal is generated and added to the input audio signal from the microphone 2 to remove the howling component. Therefore, the adaptive type as in the first and second embodiments is used. Compared with the howling canceller 7, the howling component can be quickly removed.

  The embodiment of the present invention can employ the following modifications.

  (1) In the first and third embodiments, the loudspeaker 1 is configured so that the microphone 2 and the speaker 3 can be attached externally. Good.

  (2) In the first embodiment, a configuration is used in which a signal in the frequency band detected from the audio signal x (k) is extracted by the bandpass filter 71 and input to the delay unit 72 as the audio signal x ′ (k). However, it is not limited to this configuration. For example, the band pass filter 71 may be arranged on the path of the residual signal e (k) from the subtraction unit 74 to the adaptive filter 73. In this configuration, the audio signal x (k−τ) is input to the adaptive filter 73. Then, the detected frequency band signal is extracted from the residual signal e (k) and input to the adaptive filter 73 as a reference signal. The adaptive filter 73 calculates filter coefficients based on the audio signal x (k−τ) and the reference signal. Even with such a configuration, it is possible to achieve the object of the first embodiment in which the adaptive filter 73 is identified by the transfer function h within the detected frequency band. In addition, the method of the first embodiment in which the audio signal x ′ (k−τ) is input to the adaptive filter 73 and the frequency band signal detected from the residual signal e (k) are extracted to the adaptive filter 73. The object of the first embodiment can also be achieved by a configuration in which the above-described input method is used in combination.

  (3) In the first embodiment, the adaptive filter 73 is only capable of executing an adaptive algorithm in the time domain, but a frequency domain type adaptive filter may be used instead. In the frequency domain type adaptive filter, an adaptive algorithm can be executed for each frequency band frequency-analyzed by the howling frequency detector 6. Then, the frequency band detected by the howling frequency detector 6 is notified to the frequency domain type adaptive filter, the filter coefficient is updated only in the notified frequency band, and the filter coefficient is updated in the other frequency bands. Not done. As a specific method for updating the filter coefficient only in the notified frequency band, for example, the value of the step size outside the detected frequency band is set to zero. Here, since the filter coefficient is obtained by (correction amount × step size value) + (forgetting coefficient × filter coefficient before update), the filter coefficient is calculated by setting the step size value to 0. Disappear. As a result, the amount of calculation for calculating the filter coefficient can be reduced, and the transfer function h can be identified by the transfer function h in the range of the frequency band detected using simpler processing.

It is a block diagram which shows schematic structure of the loudspeaker concerning 1st Embodiment. It is a block diagram which shows schematic structure of the loudspeaker concerning 3rd Embodiment.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1,1A, 1B-Loudspeaker 2-Microphone 3-Speaker 5-Power amplifier 6, 6B- Howling frequency detection part 7, 10- Howling canceller 71- Band pass filter 73- Adaptive filter 74- Subtraction part 11- Cancellation signal generation Unit 12-adder unit 100-feedback transmission system

Claims (4)

Connect the speaker and microphone,
A loudspeaker that loudspeaks an input audio signal input from the microphone and supplies the loudspeaker to the speaker;
A simulation signal generating unit that generates a simulation signal for simulating feedback sound returned from the speaker to the microphone;
A loudspeaker comprising: a subtracting unit that generates a residual signal obtained by subtracting the simulated signal generated by the simulated signal generating unit from an input audio signal from the microphone, and outputs the residual signal to the loudspeaker unit. ,
A howling frequency detection unit for detecting a frequency of a component of a howling sound included in an input sound signal from the microphone;
A band limiting unit that limits the generation of a simulation signal that simulates feedback sound only in a frequency band in the vicinity of this frequency including the frequency detected by the howling frequency detection unit;
A loudspeaker comprising: The band limiting unit is a bandpass filter that allows only a component in a frequency band in the vicinity of this frequency including a frequency detected by the howling frequency detection unit among audio signals input to the speaker to pass.
2. The simulation signal generation unit includes an adaptive filter that generates a simulation signal using a residual signal input from the subtraction unit and an audio signal that has passed through the bandpass filter. The loudspeaker described. Connect the speaker and microphone,
A loudspeaker that loudspeaks an input audio signal input from the microphone and supplies the loudspeaker to the speaker;
A simulation signal generating unit that generates a simulation signal for simulating feedback sound returned from the speaker to the microphone;
A loudspeaker comprising: a subtracting unit that generates a residual signal obtained by subtracting the simulated signal generated by the simulated signal generating unit from an input audio signal from the microphone, and outputs the residual signal to the loudspeaker unit. ,
A howling frequency detection unit for detecting a frequency of a component of a howling sound included in an input sound signal from the microphone;
A band limiting unit that limits the subtracting unit to subtract only the frequency band components in the vicinity of this frequency including the frequency detected by the howling frequency detecting unit of the simulated signal;
A loudspeaker comprising: Connect the speaker and microphone,
A loudspeaker device comprising a loudspeaker that loudspeaks an input audio signal input from the microphone and supplies it to the speaker,
A howling frequency detection unit that detects the frequency and phase of howling sound components included in the input sound signal from the microphone;
A cancellation signal generation unit that generates a cancellation signal that is the same frequency as the frequency detected by the howling frequency detection unit and that is opposite in phase to the detected phase;
An adder that adds the cancellation signal generated by the cancellation signal generator to the input audio signal from the microphone;
A loudspeaker comprising:

Images