JP4573689B2 - Howling prevention device - Google Patents

Description

  The present invention relates to a howling prevention device in an acoustic device.

  In an audio device that amplifies an audio signal input from a microphone and outputs it from a speaker, howling may occur if the microphone is brought close to the speaker or the output level from the speaker is increased. This is because the sound output from the speaker is fed back to the microphone and enters an oscillation state.

In order to prevent this howling, many proposals have been made in the past. For example, in Japanese Patent No. 2773656, white noise is output as a test signal from a speaker to a space, and the sound in the space is input by a microphone. An anti-howling device is disclosed in which howling is measured by measuring the frequency characteristic of the filter, determining the frequency characteristic of the filter based on the measurement result, and lowering the level of a specific frequency.
Japanese Patent No. 2773656

  However, the above-described howling prevention device has a problem that white noise is generated as a test signal, which is very annoying to humans. In addition, when using a sine wave as a test signal, it is similarly harsh and must generate a large number of test signals covering the entire audible frequency band, and the test signals are generated by sequentially changing the frequency. It takes time to do so and has the disadvantages of complicated processing.

  The present invention has been made to solve the above-described problems, and an object of the present invention is to provide a howling prevention apparatus that generates a test signal that is harsh to humans with a simple configuration.

In order to achieve this object, a howling prevention apparatus according to claim 1 detects an input means for inputting an audio signal, and a frequency characteristic of the audio signal input by the input means, and converts the detected frequency characteristic into the detected frequency characteristic. Filter means for lowering the output level of a specific frequency component of the audio signal input by the input means in response, and a sample point having a series of amplitude values sampled at a predetermined sampling period ; Sound source means for generating an audio signal having a sample point with an amplitude value of 0 at a sampling period obtained by dividing the predetermined sampling period into a plurality of equal parts between the adjacent sample points having amplitude values, and the sound source means Output means for outputting the sound signal generated by the space to the space, and the input means inputs the sound signal output by the output means. That.

Howling preventing apparatus according to claim 2, wherein, in the howling preventing apparatus according to claim 1, wherein said sound source means includes a waveform storage means for storing a set of amplitude values that have been sampled a predetermined waveform at the predetermined sampling period, Between the amplitude value reading means for sequentially reading the amplitude value from the waveform storage means and the adjacent amplitude value read by the amplitude value reading means, the amplitude value is 0 in a sampling period obtained by dividing the predetermined sampling period into a plurality of equal parts. And a sample point adding means for adding a sample point.

  The howling prevention apparatus according to claim 3 is the howling prevention apparatus according to claim 2, wherein the waveform storage means stores a waveform of an audio signal in a predetermined frequency band, and the sound source means includes the amplitude value. Pitch change means is provided for controlling to read the waveform by changing the pitch of the waveform read by the reading means.

According to the howling prevention apparatus of the first aspect, the input means for inputting the audio signal, and the frequency characteristic of the audio signal input by the input means are detected, and the input means is used according to the detected frequency characteristic. be those provided with filter means for reducing the output level of a specific frequency component of the input audio signal, the sound source means includes a sample point having a series of amplitude values sampled at a predetermined sampling period, next the An audio signal having a sampling point with an amplitude value of 0 at a sampling period obtained by dividing the predetermined sampling period into a plurality of equal parts between sample points having matching amplitude values is generated. Since the generated audio signal is output to the space, the sound emitted into the space is different from a simple sine wave or white noise, and has a sound with an appropriate overtone. Because, there is an effect that a voice is not harsh by people.

According to the howling preventing apparatus according to claim 2, in addition to the effects of the howling preventing apparatus according to claim 1, the sound source means stores a set of amplitude values that have been sampled a predetermined waveform at the predetermined sampling period a waveform storage means, and reading the amplitude value reading means sequentially amplitude values from the waveform storage means, between the amplitude values adjacent its amplitude value reading means has read, sampling period of the predetermined sampling period to more equally And a sample point adding means for adding a sample point with an amplitude value of 0, there is an effect that a test signal with good harshness can be formed with a simple configuration.

  Further, only the amplitude value of the waveform is stored in the waveform storage means, and sample points with an amplitude value of 0 are additionally inserted by the sample point addition means, so the storage capacity of the waveform storage means can be reduced. .

  According to the howling prevention apparatus of the third aspect, in addition to the effect of the howling prevention apparatus according to the second aspect, the waveform storage means stores the waveform of the audio signal in a predetermined frequency band, and the sound source means Since the pitch value changing means for changing the pitch of the waveform read by the amplitude value reading means and controlling it to be read is provided, it is possible to form a harsh sound over the audible band. is there.

  Further, compared to the case where a test signal that covers the entire audible frequency is generated by a simple sine wave, the sound source means of the present invention generates a sound containing a lot of overtones. The number of waveforms can be reduced, the time required to generate the test signal is short, and the processing can be simplified.

  Hereinafter, preferred embodiments of the present invention will be described with reference to the accompanying drawings. FIG. 1 is a block diagram showing an electrical configuration of a howling apparatus 1 according to the present invention. As shown in FIG. 1, a howling prevention apparatus 1 includes a CPU 2, a ROM 3 for storing a program executed by the CPU 2, a RAM 4, an operation panel 5, and a DSP 9 connected to each other via a bus. The microphone 6, the microphone amplifier 7, and the A / D converter 8 are connected to the input side, and the D / A converter 10 and the D / A converter 10 convert the output side. An amplifier 11 for amplifying an analog signal and a speaker 12 are provided.

FIG. 2 is an operation panel diagram showing details of the operation panel 5. The operation panel 2 includes a power switch 5a for turning on / off the power of the howling prevention device 1 and a plurality of operators corresponding to each channel. It has been. This section describes the operation elements are found with the channel 1, the operator of the other channels will be omitted.

  The input select switch 5b is a switch for setting whether the input device connected to the channel 1 outputs a microphone level or a line level. When the microphone level is selected, the LED 1 When the level is selected, LED 2 is turned on.

  The reverb / delay setting knob 5c sets the depth of the effect of either reverberation or delay added to the sound input to this channel. When the knob is set to the left half position, the reverb depth is set. When the knob is set to the right half position, the delay depth is set. The volume knob 5d is a knob for adjusting the volume of the sound input to this channel.

  The anti-feedback switch 5e is a switch for switching between enabling and disabling the anti-feedback function (howling prevention) of this channel, and the LED 3 is turned on when the anti-feedback function is set to be valid. When the anti-feedback function is set to be valid, the level of a predetermined frequency of the input audio signal is lowered by a notch filter unit 23 (see FIG. 3) described later.

  The scan switch 5f is a switch that activates a built-in sound source in order to set the frequency and the like of the notch filter unit 23 when the anti-feedback function is valid. The LED 4 blinks until the sound source starts oscillating and the setting of the notch filter is completed.

  FIG. 3 is a functional block diagram of the DSP 9. Note that functions relating to effects such as reverb and delay are omitted. The digital amplitude value converted at a predetermined sampling frequency (for example, 48 kHz) by the A / D converter 8 is input to the band dividing unit 21 and the notch filter unit 23.

  The band dividing unit 21 is composed of a bandpass filter that divides a range of 20 Hz to 20 kHz, which is a band of input voice, into 50 bands, and the peak detection unit 22 performs a level peak for each divided band. A value is detected. The peak detection circuit corresponding to each band of the peak detection unit 20 is configured such that the peak value is set to 0 at the time of starting the measurement, and thereafter holds the largest value of the input level.

  According to the peak value of each band detected by the peak detector 22, a frequency with a high probability of howling is detected, and the notch filter unit 23 is set for the detected frequency. Details of which frequency is selected will be described later.

  The DSP 9 is provided with a sound source unit 24, and when the scan switch 5f is operated, musical sounds having frequency components from 50 Hz to a little less than 24 kHz that cover the audible frequency band are sequentially generated at a uniform level. In the sound source unit 24, a 200 Hz sine wave is sampled at a predetermined sampling frequency Fs (48 kHz) for one period in the waveform memory 24a, and amplitude values thereof are stored in the order of addresses.

  The amplitude value reading unit 24b performs a process of sequentially reading stored amplitude values from the waveform memory 24a, and has a function of reading amplitude values one by one or every other plural number. When amplitude values are read one by one, the pitch of the read waveform is 200 Hz, and when every two amplitude values are read, the pitch of the read waveform is 600 Hz. . Similarly, the pitch of the waveform when read every four is 1000 Hz, and the pitch of the waveform when read every six is 1400 Hz.

  When the amplitude value reading unit 24b reads and outputs the amplitude value once in the period SP defined by the sampling frequency, the sampling point adding unit 24c sets the amplitude value to 0 in the next three cycles. A process for outputting a certain sample point is performed.

  The output of the sound source unit 24 and the output of the notch filter unit 23 are connected to the inputs 1 and 2 of the switch SW1, and either one is selected and output from the DSP 9 to the D / A converter 10.

  When the scan switch 5f is operated, the SW1 is connected to the input 1, and the sound generated by the sound source unit 24 is output. In this state, the sound input from the microphone 6 is band-divided by the band dividing unit 21. Then, the peak detection unit 22 detects the peak value for each band, and the notch filter unit 23 is set.

  When the setting of the notch filter unit 23 is completed, the SW 1 is connected to the input 2, the level of the frequency that is more likely to cause howling than the notch filter unit 23 is reduced, and the occurrence of howling is suppressed.

  FIG. 4 is a diagram schematically showing frequency characteristics in the space when sound having a uniform level is generated in the entire audible frequency band in the space in which the howling prevention device 1 is installed. FIG. 4 (a) is a graph showing the frequency characteristics when a person is present near the microphone (solid line) and when no person is present (dashed line), and in each case, the horizontal axis is frequency and the vertical axis is Shown as a level.

  As shown in this figure, the frequency characteristic when there is a person near the microphone approximates the frequency characteristic when there is no person near the microphone to the frequency characteristic in which the level is increased as the frequency is higher. . Therefore, it can be seen that the frequency characteristic when a person is present near the microphone and the frequency characteristic obtained by removing a gradual level change from the frequency characteristic when a person is not present near the microphone are substantially the same. .

  FIG. 4B removes a gradual level change from the frequency characteristics when a person is present near the microphone shown in FIG. 4A or the frequency characteristics when a person is not present near the microphone. Frequency characteristics. This frequency characteristic is obtained by passing the frequency characteristic shown in FIG. 4A through a high-pass filter using the level value corresponding to the frequency as a sequence. As a specific method, it can be obtained by calculating a gradual level change with respect to the frequency as a moving average, and subtracting the level value obtained by the moving average from the level value.

  FIG. 5 is a waveform diagram showing the waveform of the sound generated by the sound source unit 24. FIG. 5A shows an example of a waveform stored in the waveform memory 24a provided in the sound source unit 24. The horizontal axis indicates time t, the vertical axis indicates the amplitude value of the waveform, and a predetermined frequency characteristic. Is sampled at a predetermined sampling period SP to show a series of amplitude values a, b, c. These amplitude values are stored in the waveform memory 24a in the order of addresses.

In FIG. 5B, amplitude values are sequentially read out from the waveform memory 24a by the amplitude value reading unit 24b, and a plurality of amplitude values 0 have a predetermined sampling period SP between adjacent amplitude values by the sample point adding unit 24c. It shows a waveform inserted at a plurality of equally divided sampling periods. In this embodiment, a case where three sample points having an amplitude value of 0 are inserted is shown.

  FIG. 5C is a diagram showing the frequency characteristics of the waveform thus formed and the original waveform. In this figure, the spectrum shown as the spectrum of the original waveform (frequency fo) is a spectrum when the waveforms stored in the waveform memory are sequentially formed at the sampling period SP, and the same sampling is performed between the sampling points of this waveform. The spectrum of a waveform (hereinafter referred to as a zero-point insertion waveform) with three zero amplitude points added in the period is a frequency (fo / 4) of ¼ of the frequency fo of the spectrum of the original waveform and its alias spectrum. It is.

  As shown in the figure, the frequency of the spectrum of the zero point insertion waveform is 0 to (Fs / 8), (Fs / 8) to (Fs / 4), (Fs / 4) to (3Fs / 8), ( One is generated in each band of 3Fs / 8) to (Fs / 2), and these are generated symmetrically with respect to the (Fs / 8) axis and the (Fs / 4) axis. That is, the frequency of these spectra is fo / 4 in the range of 0 to (Fs / 8), and (Fs / 4) − (fo / in the range of (Fs / 8) to (Fs / 4). 4), and in the range of (Fs / 4) to (3Fs / 8), it is (Fs / 4) + (fo / 4), and in the range of (3Fs / 8) to (Fs / 2), ( Fs / 2)-(fo / 4).

  Therefore, when the spectrum of the waveform read from the original waveform by the amplitude value reading unit 24b changes in the range of 200 Hz to (Fs / 2), the spectrum of the zero point insertion waveform also changes in the above four bands.

  In this way, the spectrum of the original waveform is concentrated at a specific frequency, whereas the spectrum of the newly formed waveform is dispersed in the above four frequency bands due to aliasing. The effect is easy to hear and less irritating for humans.

  When the pitch of the original waveform is generated every 400 Hz as described above, a spectrum is generated every 100 Hz in each band. Therefore, a large number of different frequencies can be generated with a small pitch change.

  FIG. 6 is a flowchart showing filter setting processing in the DSP 9. This filter setting process is started when the scan switch 5f is operated. The CPU 2 detects whether or not the scan switch 5f is operated, and detects that the scan switch 5f is operated. In this case, the DSP 9 is instructed to execute the filtering process, and the DSP 9 starts executing. First, the peak value of the peak detection circuit provided for each band of the peak detector 22 is set to 0 (S1).

  Next, the interval for reading from the waveform memory 24a is set to 1 (S2). Thus, the amplitude value reading unit 24b sequentially reads the amplitude value of the 200 Hz sine wave stored in the waveform memory 24a. The sample point adding unit 24c inserts a sample point having an amplitude value of 0 in the next three sampling periods SP after the read amplitude value. As a result, the zero point insertion waveform formed includes a 50 Hz spectrum. In this way, the sound source unit starts generating sound and generates a musical sound with the same pitch for one second.

  Next, every two addresses to be read out are skipped (two amplitude values are skipped) (S3). As a result, the pitch of the original waveform read from the waveform memory is 600 Hz. In this case as well, sample points with an amplitude value of 0 are inserted in the next three sampling periods of the read amplitude value. As a result, the zero point insertion waveform includes a spectrum of 150 Hz.

  Similarly, if sound is generated at this pitch for 1 second, then every four read addresses (4 amplitude values skipped), the pitch of the read original waveform is 1000 Hz, and the zero point waveform is , Including a spectrum of 250 Hz. Thereafter, in the process of S3, the original waveform is formed by skipping even-numbered amplitude values sequentially, such as every 6th and every 8th. When the pitch is sequentially increased in this way, the pitch of the original waveform read out 60 times becomes 23.8 kHz, and a musical tone that sufficiently covers the audible band can be generated (S4).

  When the generation of the musical sound in the entire audible band is finished (S4: Yes), the next time until the sound reverberation in the space is stabilized is waited (S5). The time during which the reverberation stabilizes is about several seconds, although it varies depending on the size of the space. During this time, the band dividing unit 21 divides the band of the voice that is always input, and the peak detecting unit 22 detects and holds the peak value of the level of the input voice for each band.

  As a result, the peak value for each band is acquired. Next, for the acquired peak value, a moving average value is sequentially obtained and subtracted from the peak value. When the peak values of the respective bands divided by the band dividing unit 21 are P1 to P50, nine peak values from P (N−4) to P (N + 4) are added for the Nth peak value PN. Then, the sum is divided by 9 to obtain an average value, and the average value is subtracted from PN to obtain a corrected PN value (S6).

  By performing the calculation in this way, a peak value from which a gradual change of the peak value is removed is obtained, and a level for each band is obtained regardless of whether a person is close to the microphone. Therefore, the center frequency of the notch filter is set to a frequency having a large corrected peak value obtained in this way (S7).

  Generally, when a person is present in the vicinity of a microphone, the high frequency range of the frequency characteristic tends to be higher than when no person is present. Therefore, by raising the high frequency range of the frequency characteristic detected in the state where no person is present near the microphone, it is possible to obtain the frequency characteristic when a person is present near the microphone. Therefore, a gentle change in the frequency characteristic may be removed by raising the high frequency range of the frequency characteristic when no person is present near the microphone.

  As described above, according to the howling prevention apparatus of the present invention, since a plurality of sample points having an amplitude value of 0 in a predetermined sampling period are inserted after a sample point having an amplitude value, a spectrum serving as a fundamental tone And overtones are formed, and test sounds that are not harsh to humans are formed.

  In addition, since the waveform memory stores a waveform having an amplitude value, and a sample point having an amplitude value of 0 is inserted by the sample point adding unit, the waveform memory can store the waveform with a small capacity. .

  Although the present invention has been described based on the embodiments, the present invention is not limited to the above-described embodiments, and various improvements and modifications can be easily made without departing from the spirit of the present invention. Can be inferred.

  For example, in the above-described embodiment, the sound source is assumed to generate sound sequentially by increasing the pitch every 100 Hz. However, the range in which the sound pitch is changed is logarithmically constant, such as 100 cents or 200 cents. You may make it change with.

  In the above embodiment, the input sound is passed through a plurality of bandpass filters, and the frequency characteristics are detected by detecting the level of each band. However, the input sound is Fourier-transformed by using FFT or the like. It is good also as what detects by performing conversion.

It is a block diagram which shows the electric constitution of the howling prevention apparatus in embodiment by this invention. It is a top view which shows the operation panel of a howling prevention apparatus. It is a block diagram which shows the process in DSP functionally. It is a graph which shows the frequency characteristic of the sound in the space where a howling device is installed. It is a wave form diagram which shows the mode of the waveform produced | generated in a sound source. It is a flowchart which shows the process in DSP.

1 Howling prevention device 6 Microphone (input means)
9 DSP
12 Speaker (output means)
21 Band division unit 22 Peak detection unit 23 Notch filter unit (filter means) 24 Sound source 24a Waveform memory (waveform storage means)
24b Amplitude value reading unit (amplitude value reading means)
24c Sample point adding unit (sample point adding means)

Claims (3)

An input means for inputting an audio signal;
Filter means for detecting a frequency characteristic of an audio signal input by the input means and reducing an output level of a specific frequency component of the audio signal input by the input means in accordance with the detected frequency characteristic; In a howling prevention device,
A sample point having a series of amplitude values sampled at a predetermined sampling period, during sample points having an amplitude value adjacent its, the predetermined sampling cycle in multiple equally divided sampling period, the amplitude value 0 Sound source means for generating an audio signal having a sample point;
Output means for outputting to the space the audio signal generated by the sound source means,
The howling prevention apparatus according to claim 1, wherein the input means inputs the audio signal output by the output means. The sound source means is
A waveform storage means for storing a set of amplitude values that have been sampled a predetermined waveform at the predetermined sampling period,
Amplitude value reading means for sequentially reading amplitude values from the waveform storage means;
Sample point adding means for adding a sample point having an amplitude value of 0 in a sampling period obtained by dividing the predetermined sampling period into a plurality of equal parts between adjacent amplitude values read by the amplitude value reading means. The howling prevention apparatus according to claim 1 . The waveform storage means stores a waveform of an audio signal in a predetermined frequency band,
3. The howling prevention apparatus according to claim 2, wherein the sound source means includes a pitch change means for controlling to change and read the pitch of the waveform read by the amplitude value reading means.

Images