JPH086594A - Noise eliminating device for bone transmitted voice - Google Patents

Abstract

PURPOSE:To provide the device in which the auditory sense of incongruity caused by tooth chattering is reduced in the voice detected by bone transmission microphone. CONSTITUTION:A noise eliminating device 3 consists of a bone transmission microphone 1 which detects the vibration of a body tissue caused by uttering and converts it to voice signals, a noise detecting section 4 which detects pulse type noises contained in the detected voice signals caused by the tooth chattering of an uttering person and a noise eliminating section 5 which eliminates the pulse type noises detected by the section 4. The section 4 extracts the waveform feature of the pulse type noises detected by the microphone 1 and detects the tooth chattering. The section 5 makes the amplitudes of detected pulse type noises small or makes them mute state or the noise waveform is replaced using waveform data before and after the noise.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a device for removing pulse-like abnormal sounds (tooth contact noise) generated by hitting upper and lower teeth of a speaker from a sound detected by a bone conduction microphone. . INDUSTRIAL APPLICABILITY The device of the present invention reduces an uncomfortable sensation and discomfort caused by the tooth-contacting sound of bone conduction voice, and is generally applicable to communication using bone conduction voice.

[0002]

2. Description of the Related Art Air vibration in the vocal tract during vocalization is vibration of body tissue centering on the skull through the vocal tract wall. The signal obtained by detecting the vibration of the body tissue by pressing the vibration sensor against a relatively thin part of the surface tissue of the body such as the forehead, cheekbone, jaw, and ear canal wall also has information as voice. .

A voice signal obtained by the vibration sensor, that is, a bone conduction microphone is less affected by external noise around the user than a voice signal obtained by a normal microphone, so that the S / N ratio is high even in a high noise environment. There is an advantage that a voice signal with a good ratio can be obtained. Therefore, it is widely used in airports and factories. In order to obtain good quality bone conduction sound using this bone conduction microphone, in order to reduce the high frequency attenuation by the surface tissue which is the mass, spring and damper system as much as possible, the microphone sensing part should have a certain strength or more. It is necessary to detect the vibration of bone tissue as faithfully as possible by pressing with the force of.

[0004]

When a speaker wearing a bone conduction microphone utters a voice, the upper and lower teeth are lightly hit during the utterance, although there are individual differences in degree and frequency. This tooth contact sound is detected when the sound is detected with a normal microphone.
Since the detected tooth contact noise is small, there is no problem. However, when using a bone conduction microphone,
The vibration caused when the upper and lower teeth hit the sound is superimposed on the vibration of the bone tissue by voice, and when this is detected by the vibration detection type bone conduction microphone, the vibration due to the tooth hit is perceived as pulse noise, and the audible feeling is heard. It becomes very annoying.

Further, as described above, when the bone conduction microphone sensitive portion is pressed with a force of a certain strength or more to try to detect the vibration of bone tissue as faithfully as possible, this tooth contact sound is also clearly detected, There is a problem that the sound becomes very uncomfortable to the listener. It is an object of the present invention to provide a device for reducing an unpleasant sensation due to a tooth-contact sound.

[0006]

In order to achieve the above object, the present invention detects a vibration of body tissue caused by utterance and converts it into a voice signal, and a speaker included in the detected voice signal. The noise detecting unit for detecting the pulse noise generated by the tooth contact and the noise removing unit for removing the pulse noise detected by the noise detecting unit constitute a noise removing device for bone conduction voice.

[0007]

The voice waveform detected by the bone conduction microphone contains pulse noise due to the tooth-contact sound having a larger amplitude than the voice signal. The noise detector extracts the waveform characteristic of the pulse noise to detect the tooth-contact sound. The noise removal unit reduces the amplitude of the detected pulse noise or makes it silent, or replaces it using waveform data before and after the noise. As a result, the influence of the tooth-contact sound from the voice waveform detected by the bone conduction microphone is reduced, and the sense of discomfort in hearing is reduced.

[0008]

DESCRIPTION OF THE PREFERRED EMBODIMENTS FIG. 1 is a block diagram showing the entire system using a noise removing device for bone conduction voice according to an embodiment of the present invention. In the figure, the audio signal detected by the bone conduction microphone 1 usually includes a tooth-contact sound A as shown in the waveform diagram of FIG. This audio signal is input to the noise removing device 3 through the microphone amplifier 2. A communication device or the like may be used instead of the microphone amplifier 2.
Further, this audio signal can take the form of either an analog signal or a digital signal.

The input signal input to the noise eliminator 3 is input to the noise detector 4 and the noise eliminator 5. The noise detection unit 4 determines whether or not pulse noise due to tooth-contact noise is present in the audio signal, and outputs the determination result to the noise removal unit 5. Various methods for this determination are prepared in the present invention. The specific configuration of the noise detection unit 4 will be described later.

When the noise detector 4 does not detect pulse noise, the noise eliminator 5 outputs the input signal as it is. When the noise detection unit 4 detects pulsed noise, the amplitude of the determined section is reduced or silenced, or the speech waveform data before and after the section is used for replacement. Various methods for removing this pulse noise are also prepared in the present invention.
The specific configuration of the noise removing unit 5 will be described later.

The output signal of the noise removing device 3 is such that the amplitude of the pulse noise in the audio signal data is small or silent, or it is replaced with other data. This output signal is output to the speaker 7 through the speaker amplifier 6. The sound output from the speaker 7 is one in which the tooth-contact noise is removed because the pulse noise in the sound signal data is suppressed, and the sense of discomfort is eliminated.

A communication device or the like may be used instead of the speaker amplifier 6. Further, a recorder or the like may be used instead of the speaker 7. Next, a specific example of the noise detection unit 4 will be described. FIG. 3 is a block diagram showing a first embodiment of the noise detection unit 4. In this embodiment, the maximum amplitude value in the section (frame) having an appropriate time length of the input signal is used as the parameter representing the waveform shape.

The noise detector 4 shown in the figure comprises a frame processor 11, a maximum / minimum value detector 12, two dividers 13,
15, two delay circuits 14 and 16 and threshold value judging unit 1
It consists of 7. The operation of the noise detector 4 will be described with reference to the waveform chart of FIG. Noise detector 4
As shown in FIG. 4 (a), the input signal input to the pulse signal includes pulse noise A due to the tooth-contact noise. This input signal is input to the frame processing unit 11, and the frame processing unit 11 divides the input signal into frames of an appropriate time length and outputs the frames to the maximum / minimum value detection unit 12.

The maximum value / minimum value detection unit 12 calculates and outputs the difference between the maximum value and the minimum value in each frame for each frame. The output of this difference is directly input to the first divider 13 and also to the same first divider 13 through the first delay circuit 14. Therefore, the first divider 13 divides the difference value in the current frame and the difference value in the immediately previous frame, and the threshold value judgment unit 17
Output to.

The output of the first delay circuit 14 is the second
Is directly input to the second divider 15 and is also input to the same second divider 15 through the second delay circuit 16.
Therefore, the second divider 15 divides the difference value in the previous frame and the difference value in the previous frame, and outputs the result to the threshold value determination unit 17. The threshold determination unit 17 determines whether the larger or smaller of the two input signals or the average value thereof exceeds a predetermined threshold. When the difference exceeds a predetermined threshold value, it is determined that pulse noise, that is, tooth-contact noise has occurred in the frame, that is, the frame immediately before, and the determination result is output. FIG. 4B shows a plot of the larger of the two input signals.

According to the present embodiment, the pulse noise is determined based on the ratio of the maximum amplitude between frames, so that the noise can be accurately detected even when the amplitude of the audio signal is large or small. FIG. 5 is a block diagram showing a second embodiment of the noise detection unit. In this embodiment, the amplitude value of the input signal is directly used as the parameter representing the waveform shape.

The noise detector 4 shown in the figure is an amplitude normalizer 2
1. A smoothing filter 22 including a bandpass filter and a threshold value determining unit 23. The input signal is processed by the amplitude normalization unit 21 and the smoothing filter 22, and the threshold value determination unit 23 determines whether or not the amplitude value of the input signal exceeds a predetermined threshold value. Then, when the amplitude value exceeds a predetermined threshold value, it is determined that pulse noise, that is, tooth-contact noise is generated in the input signal, and the determination result is output.

According to this embodiment, the noise detecting section 4 can be constructed with a relatively simple circuit. Figure 6
The third embodiment of the noise detecting section 4 is shown, which is configured almost the same as the circuit of the second embodiment shown in FIG. 5, except that a square circuit is provided between the amplitude normalizing section 21 and the smoothing filter 22. 24
Is inserted. In this circuit, the square value (power) of the amplitude
Is used. The operation of the other parts is almost the same as that of the second embodiment described above, and therefore, the duplicated description here is omitted.

The noise detectors 4 shown in the first to third embodiments can be implemented independently, but by using a plurality of them in combination, it is possible to detect the tooth-contact noise more accurately. it can. For example, it may be determined that there is noise in the section determined by both the noise detection units 4 of the first and second embodiments having different thresholds.

Next, a specific example of the noise removing section 5 will be described. FIG. 7 is a block diagram showing a first embodiment of the noise removing section 5. The illustrated noise removing unit 5 is composed of a delay circuit 31 and a gain adjusting circuit 32. The operation of the noise removing unit 5 will be described with reference to the waveform chart of FIG.

The input signal input to the noise removing section 5 is
As shown in FIG. 8A, the pulse noise A is included.
In FIG. 8, the left side shows the entire waveform, and the right side shows an enlarged waveform near the pulse noise A. The gain adjustment circuit 32 does not operate when there is no detection signal from the noise detection unit 4. Then, when the noise detection unit 4 determines that pulse noise is generated and the detection signal is input, the gain adjustment circuit 32 sets the gain to 0.
And At this time, the delay circuit 31 ensures that the input signal input to the gain adjusting circuit 32 is exactly the noise portion. Therefore, in the audio signal output from the noise removing unit 5, the noise portion becomes 0 level.

The audio signal after this processing is shown in FIG. A silent section B is generated in the processed voice. This is perceptually perceived, but the discomfort due to the tooth-contact sound is considerably reduced. The gain in the section determined to be pulse noise in the gain adjusting circuit 32 is not always 0.
The gain is not limited to the above, and a certain amount of gain may be used as long as the discomfort due to the tooth-contact sound is not felt.

FIG. 9 is a block diagram showing a second embodiment of the noise removing section 5. In the illustrated noise eliminator 5, the input signal passes through the first delay circuit 33 and the switch 34.
Is input to one of the input terminals. When there is no detection signal from the noise detection unit 4, the switch 34 puts its contact point at the first position.
Is switched to the delay circuit side, and the input signal input to the noise removing unit 5 is output through the first delay circuit 33. Further, the switch 34 switches the contact to the other side when the detection signal is output from the noise detection unit 4.

The input signal is further input to one input side of the mixer 36. The other input side of the mixer 36 has a first
The input signal that has passed through the delay circuit 33 and the second delay circuit 35 is input. The output of the mixer 36 is the switching device 3
4 is input to the other input side. Also, this mixer 36
Are operated when the detection signal from the noise detection unit 4 is output.

In the above circuit, the switch 34 is switched to the first delay circuit 33 side when the normal detection signal from the noise detection section 4 is not output, and the first delay circuit 33 is switched. The input signal that has passed through is output from the noise canceller 5. When the detection signal is output from the noise detection unit 4, the switch 34 is switched to the output side of the mixer 36 and the opposite side. At the same time, the mixer 36 starts to operate. At this time, the current input signal and the input signal passed through the two delay circuits of the first delay circuit 33 and the second delay circuit 35 are input to the mixer 36. The mixer 36 mixes these two input signals and outputs them to the switch 34.

When the mixer 36 operates, that is, when the noise detection unit 4 outputs a pulse noise detection signal, the detection signal is delayed from the noise, so that the mixer 36 receives noise. Since the input signal of the section before the section and the input signal of the section after the noise are inputted, the mixer 36 mixes both input signals and outputs them. Therefore, the noise removing unit 5 outputs a signal in which the pulse noise of the input signal is removed and the section is replaced by the audio signals of the preceding and succeeding sections. As a result, when the sound is reproduced, the sense of discomfort is eliminated in the portion where the tooth-contact noise is removed.

The input and output signals obtained by the above circuit are shown in FIG. FIG. 10A shows an input signal.
In the output signal of FIG. 10B, the portion C corresponding to the pulse noise A of the input signal is replaced by using the waveforms of the preceding and succeeding sections. Further, in the mixer 36,
It is also possible to temporally change the mixing ratio of the preceding and following data so that the audio signal data are smoothly connected.

FIG. 11 shows a third embodiment of the noise eliminator 5, which has substantially the same configuration as the circuit of FIG. 9 of the second embodiment described above, except that the input side of the circuit of FIG. The analysis unit 37 and the synthesis unit 38 are connected to the output side and the output side, respectively. In the analysis unit 37 and the synthesis unit 38, for example, Fourier transform and inverse transform are performed to reduce the amount of information to be stored. Further, the LPC, that is, the linear prediction analysis or the analysis / synthesis by the wavelet transform is performed to encode / decode the voice. Also, the mixer 3 of FIG.
The block corresponding to 6 is replaced by the parameter interpolation circuit 39. The parameter interpolation circuit 39 interpolates the parameters obtained by Fourier transform or linear prediction.

The other circuit parts in FIG. 11 are the same as those in FIG. 9, and therefore duplicated description thereof will be omitted.

[0030]

According to the present invention, it is possible to detect and eliminate the tooth-contact sound included in the voice detected by the bone conduction microphone, so that it is possible to obtain a voice signal with less discomfort in the sense of hearing.

[Brief description of drawings]

FIG. 1 is a block diagram showing an entire bone conduction audio noise eliminator according to an embodiment of the present invention.

FIG. 2 is a waveform diagram showing a waveform of an audio signal including a tooth contact sound.

3 is a block diagram of a first embodiment of a noise detection unit in FIG.

4 is a waveform diagram showing a waveform of an audio signal in the circuit of FIG.

5 is a block diagram of a second embodiment of the noise detection unit in FIG.

6 is a block diagram of a third embodiment of the noise detection unit in FIG.

FIG. 7 is a block diagram of a first embodiment of the noise removing unit in FIG.

8 is a waveform diagram showing a waveform of an audio signal in the circuit of FIG.

9 is a block diagram of a second embodiment of the noise removing unit in FIG.

10 is a waveform diagram showing a waveform of an audio signal in the circuit of FIG.

11 is a block diagram of a third embodiment of the noise removing unit in FIG.

[Explanation of symbols]

 1 ... Bone conduction microphone 3 ... Noise removal device 4 ... Noise detection part 5 ... Noise removal part 11 ... Frame processing part 12 ... Maximum value / minimum value detection part 13, 15 ... Divider 14, 16, 31, 33, 35 ... Delay circuit 17, 23 ... Threshold value determination unit 21 ... Amplitude normalization unit 22 ... Bandpass filter 24 ... Square circuit 32 ... Gain adjustment circuit 34 ... Switching device 36 ... Mixer 37 ... Analysis unit 38 ... Synthesis unit 39 ... Parameter Interpolation circuit

Claims (8)

[Claims] 1. A bone conduction microphone that detects vibration of body tissue due to utterance and converts it into an audio signal, a noise detection unit that detects pulsed noise included in the detected audio signal and generated by tooth contact of a speaker, and And a noise removing unit for removing the pulse noise detected by the noise detecting unit from the audio signal. 2. The noise detecting section detects an amplitude value of the audio signal, and when the amplitude value exceeds a predetermined threshold value, determines that the noise is pulse noise due to tooth contact. The noise removing device for bone conduction voice according to claim 1. 3. The noise detecting section detects an amplitude value of the audio signal, and determines that the noise is a pulse noise due to tooth contact when the square value of the amplitude value exceeds a predetermined threshold value. The noise removing device for bone conduction voice according to claim 1, wherein 4. The noise detecting section detects an amplitude value of the audio signal, and when a value obtained by smoothing the amplitude value with a predetermined time constant exceeds a predetermined threshold value, pulse noise due to tooth contact is generated. 2. The bone conduction audio noise eliminator according to claim 1, wherein 5. The noise detection unit detects an amplitude value of the audio signal and divides a difference between a maximum value and a minimum value in a section of a predetermined time length by a difference value in a section before and after the section. The pulse noise due to tooth contact is determined when the divided value exceeds a predetermined threshold value.
Bone conduction voice noise removal device as described. 6. The noise removing unit divides the voice waveform into sections of a predetermined time length, and reduces the amplitude or silence of a section in which noise due to tooth-contact noise is detected. Item 1. A bone conduction audio noise eliminator according to Item 1. 7. The noise removing unit divides the voice waveform into sections of a predetermined time length, and replaces the amplitude of a section in which noise due to tooth-contact noise is detected with voice waveform data of the sections before and after the section. The device for removing noise of bone conduction voice according to claim 1, wherein 8. The noise removing unit divides the voice waveform into sections of a predetermined time length, and the waveform data in the section in which the noise due to the tooth-contact sound is detected is Fourier-transformed or linear-prediction of the waveform data before and after the noise. 2. The waveform data is replaced by using a parameter interpolated by a parameter obtained by the speech analysis and synthesis method according to claim 1.
Bone conduction voice noise removal device as described.