JPH03212697A - Signal processor - Google Patents

Abstract

PURPOSE:To exactly remove noise and to obtain a clear voice output by adopting the constitution to execute the input signal superposed with the noise on voice to the discrimination of the presence or absence of the voice in a voice discriminating section by using the signal from which the noise is previously removed in a canceling section. CONSTITUTION:The canceling section 2 to which the voice/noise input superposed with the noise on the voice is supplied removes the noise according to the noise prediction signal from a noise predicting section 1 and delivers the voice output. The voice discriminating section 3 supplied with this voice output delivers a voice detection output by detecting the presence or absence of the voice. The noise predicting section 3 operates to predict the noise in the voice/ noise input signal by using the voice detection output of the voice discriminating section 3 as a control signal and to supply this noise prediction signal to the canceling section 2. The presence or absence of the voice is exactly discriminated in this way regardless of the noise and the noise prediction is exactly executed. The noise removal is thus executed more effectively and the clear voice output is obtd.

Description

DETAILED DESCRIPTION OF THE INVENTION Field of Industrial Application The present invention relates to a signal processing device used for audio processing.

Prior Art FIG. 3 is a block diagram of a conventional signal processing device. As shown in the figure, an input signal in which noise is superimposed on speech is sent to a speech discriminator 9, a noise predictor 10, and a canceler 11.
and will be supplied. The voice discrimination section 9 discriminates the presence or absence of voice from the input signal and supplies the discrimination result to the noise prediction section 1o. The noise prediction section 1o predicts noise using an input signal in which noise is superimposed on speech and a signal supplied from the speech discrimination section 9 indicating the presence or absence of speech in the input signal. The canceling unit 11 removes the predicted noise signal obtained by the noise predicting unit 10 from the input signal in which noise is superimposed on the voice to obtain the output of the voice signal.

Problems to be Solved by the Invention However, in such a conventional signal processing device, when a lot of noise is superimposed on the input of the speech discrimination section 9, it is difficult to discriminate whether or not there is correct speech. Furthermore, if the presence or absence of voice is not accurately determined by the voice discrimination section 9, the correct noise component will not be predicted by the noise prediction section 1o, and the cancellation section 1
There is a problem that good audio output cannot be obtained from 1.

The present invention has been made to solve the above-mentioned problems, and an object of the present invention is to provide a signal processing device that can perform good noise removal even when an input signal has many noise components.

Means for Solving the Problems In order to achieve the above object, the present invention includes a voice discriminating section that detects the presence or absence of a voice signal from an input signal and outputs a control signal corresponding to the presence or absence of the voice/noise signal input. and a noise prediction unit that outputs the control signal and a noise prediction signal,
The present invention has a canceling section which receives the voice/noise signal and the noise prediction signal and obtains a voice output from which noise has been removed, and uses the voice output as an input signal to the voice discrimination section.

According to the above-described configuration, the present invention inputs the speech/noise signal to the cancellation section, removes the noise based on the noise prediction signal obtained by the noise prediction section in the cancellation section, and supplies it to the speech discrimination section. It is also used as audio output. The voice discrimination section operates to detect the presence or absence of voice from the signal supplied from the cancellation section and supply the result as a control signal to the noise prediction section. Therefore, the noise is removed from the voice/noise signal beforehand and input to the voice discriminator, so it is possible to accurately determine the presence or absence of voice, and it is also possible to predict noise correctly and obtain good voice output. .

EXAMPLE An example of the present invention will be described below with reference to FIG.

FIG. 1 shows a block diagram of a signal processing device in an embodiment of the present invention. As shown in the figure, the noise prediction unit 1
predicts noise based on the voice/noise input and the control signal supplied by the voice discrimination section 3 and supplies a noise prediction signal to the canceling section 2 .

The cancellation unit 2 removes noise according to the noise prediction signal supplied from the noise prediction unit 1 from the voice/noise input,
It obtains a voice output and supplies the voice output to the voice discrimination section 3. The voice discrimination section 3 is configured to detect the presence or absence of actual voice from the voice output, output the voice detection output, and supply the voice detection output to the noise prediction section 1 as a control signal.

The operation in the above configuration will be explained. The speech/noise input in which noise is superimposed on the speech is supplied to the canceling section 2, where the noise is removed according to the noise prediction signal supplied from the noise prediction section 1, and a speech output is obtained. The speech/noise input from which noise has been removed by the cancellation section 2 is then supplied to the speech discrimination section 3, which detects the presence or absence of speech and obtains a speech detection output. Here, the noise prediction unit 1 predicts the noise of the voice/noise input signal using the voice detection output indicating the presence or absence of voice supplied from the voice discrimination unit 3 as a suppression signal, and transmits the noise prediction signal to the cancellation unit 2. It operates to supply.

As described above, according to the signal processing device of the embodiment of the present invention, by performing voice discrimination from a signal from which noise has been removed from the voice/noise input, it is possible to accurately determine the presence or absence of voice regardless of noise. becomes. Also, this allows
Noise prediction can also be performed accurately, noise can be more effectively removed from voice/noise input, and clear voice output can be obtained.

Next, a second embodiment of the present invention will be described with reference to FIG.

FIG. 2 is a block diagram of a signal processing device in another embodiment of the present invention. As shown in the figure, the first canceling unit 6 removes the noise predicted by the first noise predicting unit 4 from the voice/noise input, and converts the signal into a voice discriminating unit 6. Second noise prediction unit 7. and is supplied to the second canceling unit 8. The voice discriminator 6 detects the presence or absence of voice in the signal supplied from the first canceller 6 and outputs a voice detection output. The voice detection output is supplied to the second noise prediction unit 7 as a control signal. The second canceling unit 8 removes the noise predicted by the second noise predicting unit from the signal supplied from the first canceling unit 5, and outputs the signal as audio output.

and a first noise prediction unit 4.

The second noise prediction unit 7 uses the control signal from the voice discrimination unit 6 to input voice/noise, respectively.

Noise prediction of the signal supplied from the first canceling unit 5 is performed. The second noise prediction section 7 then supplies the result to the second cancellation section 8, and the second cancellation section 8 is configured to output the result as audio.

To explain the operation of the above configuration, the voice/noise input in which noise is superimposed on the voice is supplied to the first canceling section 5, and the noise is reduced according to the noise prediction signal supplied from the first noise predicting section 4. Removal takes place. The first audio output from which noise has been removed in advance by the first canceling unit 6 is supplied to the second canceling unit 8, and is responsive to the second noise prediction signal supplied from the second noise predicting unit 7. Then, noise is removed accurately and audio output is obtained. Further, the first audio output from which noise has been removed in advance by the first canceling unit 6 is supplied to the audio determining unit 6 to detect the presence or absence of audio to obtain a audio detection output (control signal). Here, the first noise prediction unit 4 predicts the noise of the voice/noise input signal using the control signal indicating the presence or absence of voice supplied from the voice discrimination unit 6, and converts the first noise prediction signal into the first noise prediction signal. The signal is supplied to the cancellation unit 6 of No. 1. Further, the second noise prediction unit 7 similarly uses a control signal indicating the presence or absence of a voice supplied from the voice discrimination unit 6 to detect the first voice high intensity from which noise has been removed in advance by the first cancellation unit 5. It operates to more accurately predict noise from the signal and supply the first noise prediction signal to the second canceling unit 8. In this way, the second aspect of the present invention
According to the signal processing device of the embodiment, it is possible to accurately determine the presence or absence of voice regardless of noise, and by further accurately predicting and removing noise from the first voice output from which noise has been removed, - It becomes possible to remove non-stationary noise that has rapid fluctuations at the initial level.

Effects of the Invention As is clear from the above-described embodiments, according to the present invention, the presence or absence of speech is determined in the speech discrimination section using a signal from which noise has been removed in advance from an input signal in which noise is superimposed on speech in the cancellation section. Since the present invention is configured to perform this, it is possible to obtain the effect that the presence or absence of voice can be accurately determined regardless of the magnitude of the noise superimposed on the voice. Furthermore, by accurately detecting the presence or absence of voice, the ability to predict noise increases, noise can be removed accurately, and clear voice output can be obtained.

Furthermore, the configuration includes a second canceling section and a second noise predicting section which receive as input the audio output of the canceling section from which noise has been removed in advance, and outputs a signal from which the noise has been further removed by the second canceling section. This has the effect of making it possible to remove non-stationary noise that is low level and has rapid fluctuations.

[Brief explanation of drawings]

FIG. 1 is a block diagram of a signal processing device according to a first embodiment of the present invention, FIG. 2 is a block diagram of a signal processing device according to a second embodiment of the present invention, and FIG. 3 is a block diagram of a conventional signal processing device. It is a block diagram. 1...Noise prediction unit, 2...Cancellation unit, 3...Speech discrimination unit.

Claims (2)

[Claims] (1) A noise prediction unit that includes a voice discrimination unit that detects the presence or absence of a voice signal from an input signal and outputs a control signal corresponding to the presence or absence, and outputs a noise prediction signal based on the voice/noise signal input and the control signal. and a canceling unit that receives the voice/noise signal and the noise prediction signal and obtains a noise-free voice output, and uses the voice output as an input signal of the voice discrimination unit. (2) A voice discrimination unit that detects the presence or absence of a voice signal from an input signal and outputs a control signal corresponding to the presence or absence of the voice signal, and outputs a first noise prediction signal based on the voice/noise signal input and the control signal. a first noise prediction unit; a first cancellation unit that receives the voice/noise signal and the first noise prediction signal and obtains a first voice output from which noise is removed; A second noise prediction unit that outputs a second noise prediction signal based on the first voice output and the control signal is provided in a signal processing device that is used as an input signal of the voice discrimination unit, and the second noise prediction unit outputs a second noise prediction signal using the first voice output and the control signal. A signal processing device that receives the output and the second noise prediction signal and obtains a noise-free audio output.