JP2014216982A - Noise elimination device, noise elimination method, and noise elimination program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a noise elimination device, a noise elimination method, and a noise elimination program for effectively reducing wind noise and touch noise.SOLUTION: A noise elimination device includes: wind noise reduction means 106; touch noise reduction means 107; and wind noise/touch noise detection means 105 which detects wind noise and touch noise and selects output based on respective amplitudes of a difference signal and a sum signal of a first frequency band and a difference signal and a sum signal of a second frequency band of two voice channels.

Description

  The present invention relates to a noise removal device, a noise removal method, and a noise removal program.

  When recording ambient sound with a microphone mounted on a portable audio recording device, wind noise (hereinafter referred to as wind noise) or touch sound generated by touching the device housing (hereinafter referred to as touch noise) ) May be a problem.

Since wind noise mainly consists of frequency components of 1 KHz or less, it is often reduced by a method of removing wind noise components of low frequency components. As proposed in the previous application by the present applicant, Japanese Patent Application No. 2012-23013, as one of the techniques, the low frequency component of the difference signal between the Lch audio signal and the Rch audio signal is used as the wind noise component, and the frequency domain There is a method to reduce wind noise by subtracting. On the other hand, touch noise is impulsive noise, and since it has a wide frequency band, it is difficult to reduce it by removing a specific band. The impulsive noise part is attenuated or completely cut off. A method is used in which the period is interpolated with front and rear signals (Patent Document 1).
These two methods of reducing noise are effective for noise of interest, but if they are applied to noise that is not of interest, they are not only ineffective but may result in a loss of sound quality. For example, when the touch noise reduction method is used for wind noise, the recording target sound is attenuated over a relatively long time, and the sound quality is deteriorated. Further, even if a method for reducing wind noise is applied to touch noise, the noise reduction effect is small.
In Patent Document 2, the absolute value of the difference component signal of the low frequency component of each of the two microphones is compared with the absolute value of the sum component signal, and the absolute value of the difference component signal is the sum of the absolute values of the sum component signal. A method is disclosed in which wind noise reduction processing is performed by regarding wind noise when a larger state continues for a predetermined period or longer.

JP 2005-303681 A JP 2009-36831 A

However, if it is determined that the wind noise reduction means (HPF in Patent Document 2) is started after it is determined that the wind noise is continuous for a predetermined period or longer, the wind noise is not reduced during the predetermined period. turn into. Further, Patent Document 2 does not describe a means for reducing touch noise. Therefore, even with the method of Patent Document 2, there is a problem that the effect of reducing both wind noise and touch noise is not sufficient.

  The present invention has been made in view of the above problems, and provides a noise removal device, a noise removal method, and a program that can accurately discriminate between wind noise and touch noise and reduce both of them. Objective.

In order to solve the above-mentioned problems, the present invention comprises means described in the following 1) to 3).

That is,
1) first sound collecting means for obtaining a first sound signal;
Second sound collecting means for obtaining a second sound signal;
A first difference signal that is a difference between a first frequency band component of the first audio signal and a first frequency band component of the second audio signal, and a first sum signal that is the sum thereof A first calculating means for calculating
A second difference signal that is a difference between a component of the second frequency band of the first audio signal and a component of the second frequency band of the second audio signal, and a second sum signal that is the sum thereof A second calculating means for calculating
First noise reduction processing means for performing a first noise reduction process on the first voice signal and the second voice signal and outputting a first reduction process signal;
Second noise reduction processing means for performing a second noise reduction process on the first voice signal and the second voice signal and outputting a second reduction process signal;
Switching means for selecting and outputting the first reduction processing signal or the second reduction processing signal;
It is determined whether to select the first reduction processing signal by determining whether to select the first reduction processing signal based on the amplitude of the first difference signal and the amplitude of the first sum signal. A first control means for controlling the switching means to output the first reduction processing signal in the case of,
Second control means for controlling the second noise reduction processing means based on the amplitude of the second difference signal and the amplitude of the second sum signal;
A noise removal apparatus comprising:
2) The first frequency band component of the first audio signal of one channel of the plurality of audio signals input from the plurality of channels and the first audio signal of the second audio signal of the other one channel. Calculating a first difference signal that is a difference from a frequency band component and a first sum signal that is a sum thereof;
Calculating the amplitude of the first difference signal;
Calculating an amplitude of the first sum signal;
A second difference signal that is a difference between a component of the second frequency band of the first audio signal and a component of the second frequency band of the second audio signal, and a second sum signal that is the sum thereof Calculating steps,
Calculating an amplitude of the second difference signal;
Calculating an amplitude of the second sum signal;
Performing a first noise reduction process on the first audio signal and the second audio signal to calculate a first reduction processing signal;
Performing a second noise reduction process on the first audio signal and the second audio signal to calculate a second reduced signal;
Selecting and outputting the first reduced processing signal or the second reduced processing signal;
It is determined whether to select the first reduction processing signal by determining whether to select the first reduction processing signal based on the amplitude of the first difference signal and the amplitude of the first sum signal. If so, controlling to output the first reduction processing signal;
Controlling the second noise reduction process based on the amplitude of the second difference signal and the amplitude of the second sum signal;
A noise removal method comprising:
3) A noise removal program for causing a computer to execute a noise removal method for removing noise,
The first frequency band component of the first audio signal of one channel among the plurality of audio signals input from the plurality of channels and the first frequency band of the second audio signal of the other channel Calculating a first difference signal that is a difference from a component of and a first sum signal that is a sum thereof;
Calculating the amplitude of the first difference signal;
Calculating an amplitude of the first sum signal;
A second difference signal that is a difference between a component of the second frequency band of the first audio signal and a component of the second frequency band of the second audio signal, and a second sum signal that is the sum thereof Calculating steps,
Calculating an amplitude of the second difference signal;
Calculating an amplitude of the second sum signal;
Performing a first noise reduction process on the first audio signal and the second audio signal to calculate a first reduction processing signal;
Performing a second noise reduction process on the first audio signal and the second audio signal to calculate a second reduced signal;
Selecting and outputting the first reduced processing signal or the second reduced processing signal;
It is determined whether to select the first reduction processing signal by determining whether to select the first reduction processing signal based on the amplitude of the first difference signal and the amplitude of the first sum signal. If so, controlling to output the first reduction processing signal;
Controlling the second noise reduction process based on the amplitude of the second difference signal and the amplitude of the second sum signal;
A noise removal program comprising:

  According to the present invention, it is possible to provide a noise removal device, a noise removal method, and a noise removal program that can accurately discriminate between wind noise and touch noise and reduce both of them.

1 is a configuration diagram of a noise removal device according to Embodiment 1. FIG. FIG. 3 is a configuration diagram of wind noise / touch noise detection means according to the first embodiment. 3 is a configuration diagram of wind noise reduction means according to Embodiment 1. FIG. 4 is a configuration diagram of touch noise reduction means according to Embodiment 1. FIG. 6 is a graph showing the attenuation characteristics of the touch noise reduction means according to the first embodiment. 3 is a configuration diagram of switching means according to Embodiment 1. FIG. 6 is a configuration diagram of a noise removing device according to Embodiment 2. FIG. It is a block diagram of the wind noise detection means which concerns on Embodiment 2. FIG. It is a block diagram of the touch noise detection means which concerns on Embodiment 2.

DESCRIPTION OF EXEMPLARY EMBODIMENTS Hereinafter, preferred embodiments for implementing a noise removing device according to the invention will be described.
<Embodiment 1>
FIG. 1 is a diagram illustrating the configuration of the noise removal device according to the first embodiment. Stereo left and right channel (Lch, Rch) audio signals are A / D converted from two microphones (not shown) arranged in the case of the portable audio recording device and input from input terminals 101 and 102. . The two microphones are a first sound collecting means and a second sound collecting means, and the left and right channel sound signals are a first sound signal and a second sound signal.

The noise removal apparatus 100 includes wind noise / touch noise detection means 105, wind noise reduction means 106, touch noise reduction means 107, switching means 108, input terminal 101, input terminal 102, output terminal 103, and output terminal 104. .
The noise removal apparatus 100 may be realized by an analog circuit, a digital circuit, or the like, or may be realized by software such as a CPU (Central Processing Unit) or a DSP (Digital Signal Processor), or a combination thereof. May be.
The audio signal Lch from the left microphone is input to the input terminal 101. The audio signal Rch from the right microphone is input to the input terminal 102. The sound signal of each channel is input to the wind noise / touch noise detection means 105, the wind noise reduction means 106 as the first noise reduction processing means, and the touch noise reduction means 107 as the second noise reduction processing means. The wind noise / touch noise detection means 105 outputs a signal for controlling the touch noise reduction to the touch noise reduction means 107. Further, a control signal for switching the output signal of the wind noise reduction unit 106 and the output signal of the touch noise reduction unit 107 is output to the switching unit 108. The wind noise / touch noise detection means 105 also controls ON / OFF of the operation of the wind noise reduction means.
The output signal of the wind noise reduction means 106 becomes one input (A input) of the switching means 108. The output signal of the touch noise reduction means 107 becomes the other input (B input) of the switching means 108. In the switching means 108, the A input and the B input are switched and output.
The output of the switching unit 108 is input to an encoding unit (not shown), subjected to compression encoding and the like, and recorded on a recording medium. Alternatively, it is converted into an analog signal by a D / A conversion means (not shown) and output to a speaker or the like.
FIG. 2 is a diagram showing an internal configuration of the wind noise / touch noise detection means 105. Lch and Rch audio signals from the input terminals 101 and 102 in FIG. 1 are input from the input terminals 201 and 202. Each signal is input to a subtracter 205 and an adder 206, and subtraction and addition are performed to generate a difference signal and a sum signal.
The difference signal that is the output of the subtracter 205 is input to the band pass filters 207 and 209. The sum signal output from the adder 206 is input to the band pass filters 208 and 210. Here, the pass band of the band pass filter 207 is relatively low, and the pass band of the band pass filter 209 is higher than that of the band pass filter 207. For example, the pass band of the band pass filter 207 is 50 Hz to 300 Hz, and the pass band of the band pass filter 209 is 500 Hz to 1.5 KHz. The band pass filter 208 has the same characteristics as the band pass filter 207, and the band pass filter 210 has the same characteristics as the band pass filter 209.

A first difference signal is calculated by a combination of the subtractor 205 and the band pass filter 207. A first sum signal is calculated by a combination of the adder 206 and the band pass filter 208. The first calculation means includes a subtractor 205, a band pass filter 207, an adder 206, and a band pass filter 208. A second difference signal is calculated by a combination of the subtractor 205 and the band pass filter 209. A second sum signal is calculated by a combination of the adder 206 and the band pass filter 210. The second calculation means includes a subtractor 205, a band pass filter 209, an adder 206 and a band pass filter 210.
Setting the pass band of the band pass filter 207 to 50 Hz to 300 Hz is a band lower than the band used in the same case in the conventional example. In the conventional example, the pass band is 1 KHz or less. Even when the pass band is 50 Hz to 300 Hz, a large amount of low frequency components in the band exists for wind noise, so a signal having a relatively large amplitude is output as the filter output. On the other hand, touch noise is impulsive noise and has a wide frequency band, but the low frequency component of 50 Hz to 300 Hz is small, and the output amplitude of the band pass filter with a pass band of 50 Hz to 300 Hz is almost zero. .
The audio signal input to two microphones arranged close to each other is a signal having substantially the same low frequency component of several KHz or less, and a signal having a difference corresponding to the distance between the microphones at a frequency higher than that. When the speed of sound in the atmosphere is 340 m / sec, the wavelength of a signal with a frequency of 1 KHz is 34 cm, and when the distance between two microphones is 2 cm, a substantially equal signal waveform is input. Therefore, the difference signal is almost zero. On the other hand, the wavelength of a signal with a frequency of 10 KHz is 3.4 cm, and when a signal with this frequency comes in from the horizontal direction of the arrangement of two microphones, a signal waveform having opposite phases is input to the two microphones. . Therefore, the difference signal becomes a signal having a large amplitude.
The signal generated in the microphone due to wind noise is caused by the turbulent flow caused by the wind, and even if the microphones are close to each other, the situation of the turbulent flow is different, so the difference signal has a large amplitude even in the low frequency band. Have. The frequency band of the wind noise signal mainly includes components of 1 KHz or less, and includes frequency components of tens of Hz or less.
Touch noise is generated when a human finger or the like touches the casing, and is impulsive noise having a wide frequency band, but a low frequency component of 500 Hz or less is small. In addition, the touch noise is a signal having a different waveform even when two adjacent microphones are used. It is considered that the touch noise propagates through the housing and reaches the microphone, and a difference occurs when the vibration is transmitted to the microphone.
The output signal of the band-pass filter 207 to which the difference signal is input is substantially 0 for the audio signal, and is an output signal having a relatively large amplitude for the wind noise. For touch noise, the output signal is almost zero.
Similarly, the output signal of the band-pass filter 209 to which the difference signal is input is substantially 0 for the audio signal and is an output signal having a relatively large amplitude for the wind noise. Also, touch noise is an output signal with a relatively large amplitude.
The output signal of the band pass filter 208 to which the sum signal is input is a signal corresponding to the frequency distribution of the input signal for the audio signal, and an output signal having a relatively large amplitude for the wind noise. For touch noise, the output signal is almost zero.
Similarly, the output signal of the band pass filter 210 to which the sum signal is input is a signal corresponding to the frequency distribution of the input signal for the audio signal, and an output signal having a relatively large amplitude for the wind noise. Also, touch noise is an output signal with a relatively large amplitude.
The outputs of the band-pass filters 207, 209, 208, and 210 are input to absolute value converters 211, 212, 213, and 214, and are output as absolute values. The absolute value is used in order to facilitate comparison of amplitudes in the comparators 215 and 216. Instead, a value squared by a square calculator may be output. Moreover, you may make it output the data which smooth | blunted each signal made into absolute value on the time-axis by the low-pass filter etc. which are not shown in figure.
The output signal from the absolute value generator 211 and the output signal from the absolute value generator 213 are input to the comparator 215. Similarly, the output signal of the absolute value calculator 212 and the output signal of the absolute value calculator 214 are input to the comparator 216.
In the comparator 215, the output signal (absolute value signal related to the first difference signal) Sdiff_L of the absolute value converter 211 and the output signal (absolute value signal related to the first sum signal) Ssum_L of the absolute value converting means 213 are used. Compare. Also, the size of Sdiff_L is referred to. When the following equation (1) and equation (2) are established with a predetermined coefficient k_L (k_L is 0.5, for example) and a predetermined threshold value Vth_L,

比較器２１５は出力信号ＣＭＰ_L =１を出力し、それ以外の場合にＣＭＰ_L =０を出力する。ＣＭＰ_Lは、出力端子２０３から出力される。

The comparator 215 outputs the output signal CMP_L = 1, and otherwise outputs CMP_L = 0. CMP_L is output from the output terminal 203.

  In the comparator 216, the output signal (absolute value signal related to the second difference signal) Sdiff_H of the absolute value calculator 212 and the output signal (absolute value signal related to the second sum signal) Ssum_H of the absolute value calculator 214 are used. Compare. Further, the size of Sdiff_H is referred to. Further, the output signal CMP_L of the comparator 215 is referred to. When the following equation (3), equation (4), and equation (5) are satisfied with a predetermined coefficient k_H (k_H is 0.5, for example) and a predetermined threshold value Vth_H,

比較器２１６は出力信号ＣＭＰ_H =１を出力し、それ以外の場合にＣＭＰ_H =０を出力する。ＣＭＰ_Hは出力端子２０４から出力される。

The comparator 216 outputs the output signal CMP_H = 1, and otherwise outputs CMP_H = 0. CMP_H is output from the output terminal 204.

  The comparator 215 performs CMP_L based on the amplitude of the first difference signal that is the difference between the first audio signal and the second audio signal in the first frequency band and the amplitude of the first sum signal that is the sum thereof. Is generated. CMP_L = 1 is output when the amplitude of the first difference signal is greater than or equal to a predetermined threshold and greater than or equal to a predetermined ratio of the amplitude of the first sum signal. As described above, in the present embodiment, the first predetermined frequency band is set to the pass band 50 Hz to 300 Hz of the BPFs 207 and 208. The condition for the sum signal amplitude and the difference signal amplitude is established in this frequency band only in the case of wind noise.

  The comparator 216 performs CMP_H based on the amplitude of the second difference signal that is the difference between the first audio signal and the second audio signal in the second frequency band and the amplitude of the second sum signal that is the sum thereof. Is generated. CMP_H = 1 is output when the amplitude of the second difference signal is greater than or equal to a predetermined threshold and greater than or equal to a predetermined ratio of the amplitude of the second sum signal and the signal CMP_L = 0 from the comparator 215 Is done. As described above, in the present embodiment, the second predetermined frequency band is set to the pass band 500 Hz to 1.5 KHz of the BPFs 209 and 210. The conditions for the amplitude of the sum signal and the difference signal are satisfied in this frequency band in the case of wind noise and touch noise, but only in the case of touch noise by adding the condition of CMP_L = 0. It becomes.

  The first control means includes an absolute value generator 211, an absolute value device 213, and a comparator 215. The second control means includes an absolute value calculator 212, an absolute value generator 214, and a comparator 216.

  The wind noise / touch noise detection means 105 in FIG. 2 is configured to calculate the difference signal and the sum signal of the input signals and input the calculated difference signal and sum signal to the bandpass filter. A difference signal and a sum signal of the output of the band pass filter may be obtained by inputting to the band pass filter. In this case, the Lch audio signal from the input terminal 201 is input to the bandpass filters 207 and 209, and the Rch audio signal from the input terminal 202 is input to the bandpass filters 208 and 210. The difference signal and sum signal of the outputs of the bandpass filters 207 and 208 are calculated, and the difference signal is input to the absolute value generator 211. The sum signal is input to the absolute value generator 213. Further, a difference signal and a sum signal of the outputs of the band pass filters 209 and 210 are calculated, and the difference signal is input to the absolute value generator 212. The sum signal is input to the absolute value generator 214. Other configurations are the same as those in FIG.

  FIG. 3 is a diagram showing an internal configuration of the wind noise reduction means 106. Lch and Rch audio signals from the input terminals 101 and 102 in FIG. The wind noise reduction means 106 reduces the wind noise by subtracting the low frequency component of the difference signal between the Lch audio signal and the Rch audio signal as a wind noise component in the frequency domain.

  The Lch and Rch audio signals are input to the STFT processing means 308 and 310, respectively. Further, a difference signal between the Lch and Rch audio signals is calculated by the subtractor 306, multiplied by ½ by the constant multiplier 307, and then input to the STFT processing means 309.

  The STFT (Short Time Fourier Transform) processing means divides the input signal into frames of a predetermined length while shifting the input signal every predetermined time, performs a process of multiplying each frame by a predetermined time window, and sets the time window. FFT (Fast Fourier Transform) processing is performed on the received signal, and the phase value and the amplitude value at each frequency of each frame are output.

In the case of a 256th order FFT, 128 amplitude values | Y (t, k × f0) | (K = 0 to 127) and phase values φy (k × f0) (K = 0 to 127) are output. For Lch signal, | YL (t, k × f0) |, φyL (k × f0), for Rch signal, | YR (t, k × f0) |, φyR (k × f0), difference signal Is represented by | Ys (t, k × f0) |, φys (k × f0).
The amplitude value data in the low frequency range of the Lch signal and the difference signal are input to the coefficient multiplication / subtraction processing means 311. For example, there are 32 low-frequency data, | YL (t, k × f0) |, | Ys (t, k × f0) | (K = 0 to 31). The coefficient multiplication / subtraction processing means multiplies each data of | Ys (t, k × f0) | (K = 0 to 31) by a predetermined coefficient C. C is a value around 1.

  Next, the multiplication result is subtracted from the amplitude value data of the Lch signal. Here, when the subtraction result becomes negative as in the following equation (6), it is replaced with 0.

この減算結果とＬｃｈの信号の高周波域の振幅値データ｜ＹL(t,k×f0)｜ (K=32〜127)とＬｃｈの信号の位相値 φyL(k×f0) (K=0〜127) は、ＩＦＦＴ処理手段３１３に入力する。ＩＦＦＴ処理手段３１３では、これらの振幅情報と位相情報を用いて、ＩＦＦＴ(Inverse FFT)処理を行う。

This subtraction result and high-frequency amplitude value data of the Lch signal | YL (t, k × f0) | (K = 32 to 127) and the phase value of the Lch signal φyL (k × f0) (K = 0 to 127) ) Is input to the IFFT processing means 313. The IFFT processing means 313 performs IFFT (Inverse FFT) processing using these amplitude information and phase information.

The output of the IFFT processing unit 313 is input to the waveform synthesis unit 315. The waveform synthesizing unit 315 performs inverse wind processing and waveform synthesis processing, and outputs an Lch audio signal from the output terminal 303.
Similarly, the amplitude value data in the low frequency range of the Rch signal and the difference signal are input to the coefficient multiplication / subtraction processing means 312 and are output after being multiplied and subtracted by the coefficient. This subtraction result and the phase value φyR (k × f0) (K = 0 to 127) of the Rch signal are input to the IFFT processing means 314. The IFFT processing means 314 performs IFFT (Inverse FFT) processing using these amplitude information and phase information. The output of the IFFT processing unit 314 is input to the waveform synthesis unit 316, the same processing as that of the waveform synthesis unit 315 is performed, and an Rch audio signal is output from the output terminal 304.

  The wind noise reduction unit 106 is controlled so as to operate only when the output signal CMP_L of the comparison unit 215 is 1 and to stop operating when the CMP_L is 0. The wind noise reduction means 106 has a relatively large amount of calculation and consumes a lot of power during operation. Since wind noise does not always occur, power consumption can be reduced by stopping the operation during periods when wind noise is not occurring.

  FIG. 4 is a diagram illustrating an internal configuration of the touch noise reduction unit 107. Lch and Rch audio signals from the input terminals 101 and 102 in FIG. Further, an output signal CMP_H which is a touch noise detection result is input to the input terminal 405 from the wind noise / touch noise detection unit of FIG. The touch noise reduction unit 107 attenuates the audio signal for a predetermined period from the time when the touch noise is detected.

  When the rising edge of the input signal 405 is detected, the counter 406 is reset. Thereafter, the count-up is performed in synchronization with the sampling period of the audio signal, and when reaching a value corresponding to a predetermined time T3, the count-up is stopped and the value is held. The count value of the counter 406 is input to the attenuation curve generation means 407. The attenuation curve generation unit 407 generates an attenuation coefficient corresponding to the attenuation curve shown in FIG. 5 according to the input count value. That is, the count value corresponding to the predetermined time T1 decreases according to the count value, maintains that value until T2, and increases until T3. Specifically, it can be configured by a ROM (read only memory) or the like that outputs data related to the attenuation coefficient using data related to the input count value as an address.

  The attenuation coefficient output from the attenuation curve generation unit 407 is input to the multiplier 408 and the multiplier 409. An audio signal input from the input terminal 401 is input to the other input of the multiplier 408, and is multiplied by an attenuation coefficient and output. Similarly, the audio signal input from the input terminal 402 is input to the other input of the multiplier 409, multiplied by an attenuation coefficient, and output.

  FIG. 6 is a diagram illustrating an internal configuration of the switching unit 108. Lch and Rch audio signals, which are output signals of the wind noise reduction means 106 in FIG. 1, are input from the input terminals 501 and 502, respectively. Also, Lch and Rch audio signals, which are output signals of the touch noise reduction means 107 in FIG. 1, are input from the input terminals 503 and 504, respectively. Further, the wind noise detection signal CMP_L is input from the input terminal 505 from the wind noise / touch noise detection means of FIG. The switching unit 108 switches and outputs the input signal from the wind noise reduction unit 106 and the input signal from the touch noise reduction unit 107.

  The audio signal from the input terminal 501 is input to the coefficient multiplier 508. The audio signal from the input terminal 502 is input to the coefficient multiplier 510. The wind noise detection signal CMP_L from the input terminal 505 is input as a control signal for the coefficient multiplier 508 and the coefficient multiplier 510. The coefficient multiplier 508 and the coefficient multiplier 510 multiply the multiplication coefficient 1 when the wind noise detection signal CMP_L = 1, and multiply the multiplication coefficient 0 when CMP_L = 0.

The audio signal from the input terminal 503 is input to the coefficient multiplier 509. The audio signal from the input terminal 504 is input to the coefficient multiplier 511. The coefficient multiplier 509 and the coefficient multiplier 511 multiply the multiplication coefficient 0 when the wind noise detection signal CMP_L = 1, and multiply the multiplication coefficient 1 when CMP_L = 0.
The output of the coefficient multiplier 508 and the output of the coefficient multiplier 509 are input to the adder 512, the addition operation is performed, and the output is output from the output terminal 506. The output of the coefficient multiplier 510 and the output of the coefficient multiplier 511 are input to the adder 513, added, and output from the output terminal 507.

  Accordingly, the wind noise is detected, and the output signal of the wind noise reduction means is selected and outputted during the period when CMP_L = 1, and the output of the touch noise reduction means during the period when CMP_L = 0 without detecting the wind noise. A signal is selected and output.

  The signal CMP_L from the input terminal 505 may be input to time constant means (not shown), and the output signal α may be input to the coefficient multipliers 508, 509, 510, and 511. In the time constant means, when CMP_L rises from 0 to 1, α = 0, α = 0.2, α = 0.4, α = 0.6, α = 0. 8, Increase to α = 1.0 stepwise. When CMP_L falls from 1 to 0, α = 1.0, α = 0.8, α = 0.6, α = 0.4, α = 0.2, in synchronization with the sampling period of the audio signal. Decrease stepwise with α = 0. The output signal α is used as a multiplication coefficient in the coefficient multipliers 508 and 510, and is used as a multiplication coefficient as (1−α) in the coefficient multipliers 509 and 511. In this way, the sense of incongruity associated with switching can be reduced by changing in a staircase pattern.

  As described above, in the present embodiment, the first sum that is the sum and amplitude of the first difference signal, which is the difference between the components of the first frequency band of the first audio signal and the second audio signal, respectively. Based on the signal CMP_L generated based on the amplitude of the signal, the signal subjected to the wind noise reduction process, which is the first noise reduction process, is selected and output. Further, based on the signal CMP_L and the amplitude of the second sum signal that is the sum and the amplitude of the second difference signal that is the difference between the components of the second frequency band of the first audio signal and the second audio signal, respectively. The touch noise reduction process, which is the second noise reduction process, is controlled.

  Further, the operation of the wind noise reduction processing means 106 is stopped while the signal CMP_L is 0.

  According to the present embodiment, the wind noise reduction process is not erroneously performed on the touch noise, and the touch noise reduction process is not erroneously performed on the wind noise. As a result, reduction processing that matches the characteristics of noise can be performed, so that deterioration in sound quality can be minimized. Further, since the operation of the wind noise reduction means 106 can be stopped except when wind noise occurs, power consumption can be reduced.

Note that the comparison means 216 in the wind noise / touch noise detection means 105 in FIG. 2 may generate CMP_H without making the expression (5) a requirement. In this case, the touch noise reduction process functions even in the case of wind noise, which is disadvantageous in terms of power consumption. However, since the output signal from the wind noise reduction unit 106 is selected by the switching unit 108, the output signal is not affected.
<Embodiment 2>
Next, a second embodiment of the noise removal apparatus according to the present invention will be described with reference to the drawings.

FIG. 7 is a diagram illustrating the configuration of the noise removal device according to the second embodiment, in which components that are the same as or correspond to those in FIG. 1 according to the first embodiment are denoted by the same reference numerals.
As in the first embodiment, stereo left and right channel (Lch, Rch) audio signals are A / D converted from two microphones and input from input terminals 601 and 602. The input Lch and Rch audio signals are input to the B input of the wind noise reduction means 106, the wind noise detection means 605 and the switching means 108 which are the first noise reduction processing means. The wind noise reduction means 106 is the same as the wind noise reduction means 106 of FIG. When a delay occurs due to wind noise reduction processing with respect to the input to the wind noise reduction means 106, the delay means (not shown) is provided before the B input of the switching means 108 to compensate for the delay. You may do it.

The output signal of the wind noise reduction means 106 becomes the A input of the switching means 108. Switching means 108
These are the same as the wind noise reduction means 106 of FIG. The wind noise detection means 605 detects the presence or absence of wind noise and outputs a control signal for switching the output signal of the wind noise reduction means 106 and the signals from the input terminals 601 and 602 to the switching means 108.

  The output signal of the switching means 108 is input to the touch noise detection means 606 and the touch noise reduction means 107 that is the second noise reduction processing means. The touch noise reduction means 107 is the same as the touch noise reduction means 107 in FIG. The touch noise detecting unit 107 detects a touch noise and outputs a signal for controlling the touch noise reduction to the touch noise reducing unit 107.

FIG. 8 is a diagram showing an internal configuration of the wind noise detection means 605. Lch and Rch audio signals from the input terminals 601 and 602 in FIG. 6 are input from the input terminals 701 and 702. Each signal is input to band pass filters 706 and 707.
The band pass filters 706 and 707 are the same as the band pass filter 207 in FIG. 2, and the pass band is 50 Hz to 300 Hz. The outputs of the band pass filters 706 and 707 are input to a subtracter 704 and an adder 705, and are subtracted and added to generate a difference signal and a sum signal. The difference signal is converted into an absolute value by an absolute value calculator 708, and the sum signal is converted into an absolute value by an absolute value converter 709 and input to a comparator 710.
The first calculation means includes band pass filters 706 and 707, a subtracter 704 and an adder 705.

In the comparator 710, the output signal (absolute value signal related to the first difference signal) Sdiff_L of the absolute value calculator 708 and the output signal (absolute value signal related to the first sum signal) Ssum_L of the absolute value calculator 709 are obtained. Compare. Further, the magnitude of the absolute value signal input Sdiff_L related to the first difference signal is referred to.
When the following equation (7) and equation (8) are established with a predetermined coefficient k_L (k_L is 0.5, for example) and a predetermined threshold value Vth_L:

比較器７１０は出力端子７０３から出力信号ＣＭＰ_L =１を出力し、それ以外の場合にＣＭＰ_L =０を出力する。

The comparator 710 outputs the output signal CMP_L = 1 from the output terminal 703, and otherwise outputs CMP_L = 0.

  The first control means includes an absolute value generator 708, an absolute value device 709, and a comparator 710.

  The output signal CMP_L of the comparator 710 is input to the input terminal 504 of the switching unit 108. The internal configuration of the switching means 108 is as shown in FIG. 6 and performs the operation as described in the first embodiment.

  As described above, in the low frequency band of the audio signal from the two adjacent microphones, particularly the difference signal in the frequency band of 300 Hz or less, the audio signal and the touch noise component are almost 0, and the wind noise is the main component. . Therefore, the output CMP_L from the wind noise reduction means 106 is CMP_L = 1 only when there is wind noise, and CMP_L = 0 when there is no wind noise and there is an audio signal or touch noise.

  The switching means 108 selects and outputs the A input when CMP_L = 1, and selects and outputs the B input when CMP_L = 0. Similarly to the first embodiment, the operation of the wind noise reduction unit 106 is stopped when CMP_L = 0.

FIG. 9 is a diagram illustrating an internal configuration of the touch noise detection unit 606. The Lch and Rch audio signals from the switching means 108 in FIG. 7 are input from the input terminals 801 and 802. Each signal is input to a subtracter 804 and an adder 805, and subtraction and addition are performed to generate a difference signal and a sum signal. The difference signal is input to the band pass filter 806. The sum signal is input to the band pass filter 807. The bandpass filters 806 and 807 are the same as the bandpass filter 209 in FIG. 2, and the passband is 500 Hz to 1.5 KHz.
The second calculation means includes a subtracter 804, an adder 805, a band pass filter 806 and a band pass filter 807.

  The output of the band pass filter 806 is converted into an absolute value by the absolute value generator 808 and the output of the band pass filter 807 is converted into an absolute value by the absolute value generator 809 and input to the comparator 810.

  In the comparator 810, the output signal (absolute value signal related to the second difference signal) Sdiff_H of the absolute value calculator 808 and the output signal (absolute value signal related to the second sum signal) Ssum_H of the absolute value calculator 809 are used. Compare. Further, the magnitude of the absolute value signal Sdiff_H related to the second difference signal is referred to.

  When the following equation (9) and equation (10) are established with a predetermined coefficient k_H (k_H is 0.5, for example) and a predetermined threshold Vth_H:

比較器８１０は、出力端子８０３から出力信号ＣＭＰ_H =１を出力し、それ以外の場合にＣＭＰ_H =０を出力する。
第二の制御手段は、絶対値化器８０８、絶対値化器８０９および比較器８１０とで構成されている。

The comparator 810 outputs the output signal CMP_H = 1 from the output terminal 803, and outputs CMP_H = 0 in other cases.
The second control means includes an absolute value converter 808, an absolute value converter 809, and a comparator 810.

  The output signal CMP_H is input to the touch noise reduction unit 107. The touch noise reduction means 107 is the same as the touch noise reduction means 107 of FIG. 1 of the first embodiment, and its internal configuration is shown in FIG. CMP_H is input to the input terminal 405. The operation of the touch noise reduction unit 107 is as described above.

  The passbands of the bandpass filters 806 and 807 are 500 Hz to 1.5 KHz. As described above, when the difference signal between the two microphones arranged close to each other is input, the amplitude of the output signal for the audio signal Becomes almost zero. On the other hand, touch noise becomes an output signal having a relatively large amplitude when a difference signal is input. Originally, the wind noise also becomes an output signal having a relatively large amplitude, but in the case of the present embodiment, it is a difference signal with respect to the signal whose wind noise component has been reduced by the wind noise reduction means 106 in the previous stage. Therefore, the amplitude of the output signal due to wind noise is small.

  Accordingly, the touch noise detection unit 606 can detect only the touch noise by comparing the absolute value related to the second difference signal and the absolute value related to the second sum signal as described above. Furthermore, the touch noise can be effectively reduced because the touch noise reduction means can be operated only when the touch noise is input based on the detected signal.

The wind noise detection means 605 in FIG. 8 is configured to calculate the difference signal and the sum signal after first applying a band pass filter, but first calculates the difference signal and the sum signal as shown in FIGS. Thus, a band pass filter may be applied to them. In the touch noise detection means of FIG. 9, the difference signal and the sum signal are first calculated and a band pass filter is applied to them. However, as shown in FIG. The sum signal may be calculated.

  Further, the modification described in the first embodiment is performed in the same manner, such as replacement of the absolute value converter with a square calculator and smoothing the absolute value signal on the time axis with a low-pass filter or the like. Can do.

As described above, in the present embodiment, the first difference signal is the sum and amplitude of the first difference signal that is the difference between the components of the first predetermined frequency band of the first sound signal and the second sound signal. Based on the amplitude of the sum signal, the signal that has been subjected to the wind noise reduction process, which is the first noise reduction process, and the voice signal that has not been subjected to the process are selected and output. Further, the output signal is the sum and amplitude of the second difference signal, which is the difference between the components of the second predetermined frequency band of the signal related to the first audio signal and the signal related to the second audio signal, respectively. A second noise reduction process is performed based on the amplitude of the two sum signals.
According to the present embodiment, the wind noise reduction process is not erroneously performed on the touch noise, and the touch noise reduction process is not erroneously performed on the wind noise. As a result, reduction processing that matches the characteristics of noise can be performed, so that deterioration in sound quality can be minimized. Moreover, since the operation of the wind noise reduction means can be stopped except when wind noise occurs, power consumption can be reduced.

The above-described process for removing noise may be executed by a computer program. The noise removal program described above can be stored using various types of non-transitory computer readable media and supplied to a computer. Non-transitory computer readable media include various types of tangible storage media. Examples of non-transitory computer-readable media include magnetic recording media (for example, flexible disks, magnetic tapes, hard disk drives), magneto-optical recording media (for example, magneto-optical disks), CD-ROMs (Read Only Memory), CD-Rs, CD-R / W, semiconductor memory (for example, mask ROM, PROM (Programmable ROM), EPROM (Erasable PROM), flash ROM, RAM (Random Access Memory)) are included.
Also, the noise removal program may be supplied to the computer by various types of transitory computer readable media. Examples of transitory computer readable media include electrical signals, optical signals, and electromagnetic waves. The temporary computer-readable medium can supply the computer with a noise removal program via a wired communication path such as an electric wire and an optical fiber, or a wireless communication path.

  In addition, when the computer executes the noise removal program that implements the functions of the above-described embodiments, the noise removal program is run on the computer. A case where the functions of the above-described embodiment are realized in cooperation with an operating system (OS) or application software that is included in the embodiment is also included in the embodiment of the present invention.

100,600 Noise removal apparatus 101,102,201,202,301,302 Input terminals 401,402,405,501,502,503,504 Input terminals 601,602,701,702,801,802 Input terminals 103,104 , 203, 204, 303, 304 Output terminal 403, 404, 506, 507, 603, 604 Output terminal 703, 803 Output terminal 105 Wind noise / touch noise detection means 106 Wind noise reduction means 107 Touch noise reduction means 108 Switching means 205 Subtractor 206 Adder 207, 208, 209, 210 Band pass filter (BPF)
211, 212, 213, 214 Absolute value generator 215, 216 Comparator 306 Subtractor 307 Constant multiplier 308, 309, 310 STFT processing means 311, 312 Coefficient multiplication / subtraction processing means 313, 314 IFFT processing means 315, 316 Waveform Combining means 406 Counter 407 Attenuation curve generating means 408, 409 Multiplier 508, 509, 510, 511 Coefficient multiplier 512, 513 Adder 605 Wind noise detecting means 606 Touch noise detecting means 706, 707, 806, 807 Band pass filter ( BPF)
708, 709, 808, 809 Absolute value converter 704, 804 Subtractor 705, 805 Adder 710, 810 Comparator

Claims (7)

First sound collecting means for obtaining a first sound signal;
Second sound collecting means for obtaining a second sound signal;
A first difference signal that is a difference between a first frequency band component of the first audio signal and a first frequency band component of the second audio signal, and a first sum signal that is the sum thereof A first calculating means for calculating
A second difference signal that is a difference between a component of the second frequency band of the first audio signal and a component of the second frequency band of the second audio signal, and a second sum signal that is the sum thereof A second calculating means for calculating
First noise reduction processing means for performing a first noise reduction process on the first voice signal and the second voice signal and outputting a first reduction process signal;
Second noise reduction processing means for performing a second noise reduction process on the first voice signal and the second voice signal and outputting a second reduction process signal;
Switching means for selecting and outputting the first reduction processing signal or the second reduction processing signal;
It is determined whether to select the first reduction processing signal by determining whether to select the first reduction processing signal based on the amplitude of the first difference signal and the amplitude of the first sum signal. A first control means for controlling the switching means to output the first reduction processing signal in the case of,
Second control means for controlling the second noise reduction processing means based on the amplitude of the second difference signal and the amplitude of the second sum signal;
A noise removal apparatus comprising: The second control means is configured to control the second difference signal based on the amplitude of the first difference signal, the amplitude of the first sum signal, the amplitude of the second difference signal, and the amplitude of the second sum signal. The noise reduction apparatus according to claim 1, wherein the noise reduction processing means is controlled. First sound collecting means for obtaining a first sound signal;
Second sound collecting means for obtaining a second sound signal;
A first difference signal that is a difference between a first frequency band component of the first audio signal and a first frequency band component of the second audio signal, and a first sum signal that is the sum thereof A first calculating means for calculating
First noise reduction processing means for performing a first noise reduction process on the first voice signal and the second voice signal and outputting a first reduction process signal;
Switching means for selecting the first reduced processing signal or the first audio signal and the second audio signal and outputting the first post-selection signal and the second post-selection signal;
It is determined whether to select the first reduction processing signal by determining whether to select the first reduction processing signal based on the amplitude of the first difference signal and the amplitude of the first sum signal. A first control means for controlling the switching means to output the first reduction processing signal in the case of,
A second difference signal that is a difference between a component of the second frequency band of the first post-selection signal and a component of the second frequency band of the second post-selection signal, and a second that is a sum thereof. Means for calculating a sum signal;
Second noise reduction means for performing a second noise reduction process on the first post-selection signal and the second post-selection signal;
Second control means for controlling the second noise reduction means based on the amplitude of the second difference signal and the amplitude of the second sum signal;
A noise removal apparatus comprising: 4. The noise removal device according to claim 1, wherein the first frequency band is a frequency band lower than the second frequency band. 5. 5. The signal processing device according to claim 1, wherein the operation of the first noise reduction processing device is stopped during a period in which the first reduction processing signal is not selected. The first frequency band component of the first audio signal of one channel among the plurality of audio signals input from the plurality of channels and the first frequency band of the second audio signal of the other channel Calculating a first difference signal that is a difference from a component of and a first sum signal that is a sum thereof;
Calculating the amplitude of the first difference signal;
Calculating an amplitude of the first sum signal;
A second difference signal that is a difference between a component of the second frequency band of the first audio signal and a component of the second frequency band of the second audio signal, and a second sum signal that is the sum thereof Calculating steps,
Calculating an amplitude of the second difference signal;
Calculating an amplitude of the second sum signal;
Performing a first noise reduction process on the first audio signal and the second audio signal to calculate a first reduction processing signal;
Performing a second noise reduction process on the first audio signal and the second audio signal to calculate a second reduced signal;
Selecting and outputting the first reduced processing signal or the second reduced processing signal;
It is determined whether to select the first reduction processing signal by determining whether to select the first reduction processing signal based on the amplitude of the first difference signal and the amplitude of the first sum signal. If so, controlling to output the first reduction processing signal;
Controlling the second noise reduction process based on the amplitude of the second difference signal and the amplitude of the second sum signal;
A noise removal method comprising: A noise removal program for causing a computer to execute a noise removal method for removing noise,
The first frequency band component of the first audio signal of one channel among the plurality of audio signals input from the plurality of channels and the first frequency band of the second audio signal of the other channel Calculating a first difference signal that is a difference from a component of and a first sum signal that is a sum thereof;
Calculating the amplitude of the first difference signal;
Calculating an amplitude of the first sum signal;
A second difference signal that is a difference between a component of the second frequency band of the first audio signal and a component of the second frequency band of the second audio signal, and a second sum signal that is the sum thereof Calculating steps,
Calculating an amplitude of the second difference signal;
Calculating an amplitude of the second sum signal;
Performing a first noise reduction process on the first audio signal and the second audio signal to calculate a first reduction processing signal;
Performing a second noise reduction process on the first audio signal and the second audio signal to calculate a second reduced signal;
Selecting and outputting the first reduced processing signal or the second reduced processing signal;
It is determined whether to select the first reduction processing signal by determining whether to select the first reduction processing signal based on the amplitude of the first difference signal and the amplitude of the first sum signal. If so, controlling to output the first reduction processing signal;
Controlling the second noise reduction process based on the amplitude of the second difference signal and the amplitude of the second sum signal;
A noise removal program comprising:

Images