JP5504445B2 - Microphone device - Google Patents

Description

  The present invention relates to a microphone device, and more particularly to a microphone device using bone conduction.

  Conventionally, when humans speak, a phenomenon called bone conduction is known in which vibration of the vocal cords is transmitted through the skull and sound is transmitted directly to the auditory nerve. Since the voice transmitted through the bone conduction is hardly affected by external noise, various voice processing devices using the bone conduction have been developed in recent years, and one of them is a headset such as a mobile phone.

In a headset (microphone device) using bone conduction, a highly sensitive bone conduction microphone and an ultra-small magnetic speaker are incorporated in the headset, and has the following advantages.
(1) Two-way communication is possible simply by wearing a bone conduction microphone on one ear.
(2) Since it is difficult to pick up external noise, it has excellent noise resistance.
(3) Since the mouth is completely open, comfortable communication is possible even if a breathing apparatus that covers the face such as a gas mask is used.

  As mentioned above, the voice transmitted by bone conduction is strong against external noise in principle, but when a headset using a bone conduction microphone is used in a high environmental noise level, it is affected by the external noise. Audio quality may be degraded. Therefore, conventionally, various techniques have been proposed for making call voice clearer in such a high environmental noise level (see, for example, Patent Documents 1 and 2).

  Patent Document 1 proposes a technique for acquiring a multi-channel audio signal including a bone conduction audio signal and an air conduction audio signal, and removing a noise component from the audio signal by independent component analysis.

  Patent Document 2 discloses a bone conduction microphone worn by a user, a switch that is triggered by a signal from the bone conduction microphone and identifies a user's speech time, a normal microphone connected to the switch, There has been proposed a combination microphone system including an utterance voice analysis device provided on the output side of a switch.

  In the system of Patent Document 2, a switch is controlled by a signal from a bone conduction microphone, and a speech signal collected by a normal microphone is input to a speech speech analysis apparatus only during a user's speech time. Then, the utterance voice analysis apparatus separates the user's utterance voice and the other person's utterance voice (noise) from the input voice signal.

Japanese Patent Laid-Open No. 2007-3702 JP 2007-267331 A

As described above, conventionally, various countermeasures against external noise have been proposed in an audio processing device using a bone conduction microphone (hereinafter simply referred to as a bone conduction microphone device). However, the quality of sound transmitted by bone conduction is deteriorated not only by external noise but also by various noises generated in the user's body.

  One example of such internal noise is noise (hereinafter referred to as sudden noise) that occurs suddenly due to, for example, a tooth collision sound or a chin joint sound when a user speaks. This is internal noise that occurs at a rate of about 1 in 10 people.

  Here, in FIGS. 6A and 6B, the waveform of the audio signal detected by the bone conduction microphone when no sudden noise occurs and the voice detected by the bone conduction microphone when sudden noise occurs. Each of the signal waveforms is shown. In FIGS. 6A and 6B, the horizontal axis represents time, and the vertical axis represents signal amplitude. When sudden noise 200a is mixed during speech, when the sudden noise 200a is generated in the audio signal 200, the amplitude increases in an impulse manner as shown in FIG. 6B.

  The present invention has been made in order to solve the above-described problems, and an object of the present invention is a sudden occurrence occurring in a user's body as described above in a microphone device such as a headset using a bone conduction microphone. It is to remove noise.

In order to solve the above-described problems, a microphone device of the present invention includes a bone conduction microphone, a sudden sound extraction unit, and a sudden sound removal unit, and functions of each unit are as follows. The sudden sound extraction unit calculates an envelope of the audio signal detected by the bone conduction microphone, and extracts a signal corresponding to the signal component of the sudden noise included in the audio signal based on the envelope. The sudden sound removal unit removes the sudden noise signal component from the audio signal based on the signal corresponding to the sudden noise signal component extracted by the sudden sound extraction unit. At this time, the sudden sound extraction unit calculates an envelope of the audio signal, extracts a signal having a predetermined frequency or more from the envelope, and outputs the extracted signal as a signal corresponding to the signal component of the sudden noise.
In addition, the microphone device of the present invention includes a bone conduction microphone, a sudden sound extraction unit, and a sudden sound removal unit, and functions of each unit are as follows. The sudden sound extraction unit calculates an envelope of the audio signal detected by the bone conduction microphone, and extracts a signal corresponding to the signal component of the sudden noise included in the audio signal based on the envelope. The sudden sound removal unit removes the sudden noise signal component from the audio signal based on the signal corresponding to the sudden noise signal component extracted by the sudden sound extraction unit. At this time, the sudden sound extraction unit calculates an analysis signal of the audio signal by Hilbert transforming the audio signal at a processing frequency equal to or higher than a predetermined frequency, and calculates an envelope of the audio signal based on a real part and an imaginary part of the analysis signal. The envelope is calculated and output as a signal corresponding to the signal component of the sudden noise.

As described above, in the microphone device of the present invention, first, the sudden sound extraction unit extracts the sudden noise component signal from the envelope of the audio signal detected by the bone conduction microphone. Then, the sudden sound removal unit removes the sudden noise signal component from the audio signal based on the extracted signal corresponding to the sudden noise signal component. At this time, the sudden sound extraction unit calculates an envelope of the audio signal, extracts a signal having a predetermined frequency or more from the envelope, and outputs the extracted signal as a signal corresponding to the signal component of the sudden noise.
In addition, the microphone device of the present invention includes a bone conduction microphone, a sudden sound extraction unit, and a sudden sound removal unit, and functions of each unit are as follows. The sudden sound extraction unit calculates an envelope of the audio signal detected by the bone conduction microphone, and extracts a signal corresponding to the signal component of the sudden noise included in the audio signal based on the envelope. The sudden sound removal unit removes the sudden noise signal component from the audio signal based on the signal corresponding to the sudden noise signal component extracted by the sudden sound extraction unit. At this time, the sudden sound extraction unit calculates an analysis signal of the audio signal by Hilbert transforming the audio signal at a processing frequency equal to or higher than a predetermined frequency, and calculates an envelope of the audio signal based on a real part and an imaginary part of the analysis signal. The envelope is calculated and output as a signal corresponding to the signal component of the sudden noise.
Therefore, according to the microphone device of the present invention, the sudden noise generated in the user's body as described above can be removed from the audio signal detected by the bone conduction microphone.

1 is a schematic block configuration diagram of a bone conduction microphone device according to an embodiment of the present invention. It is a figure which shows an example of an internal structure of a sudden sound extraction part and a sudden sound removal part. It is a figure which shows the example of a waveform of an envelope signal. It is a figure for demonstrating the principle at the time of calculating an envelope signal from an audio | voice signal by the height comparison of a sampling value. It is a figure which shows one operation example of the bone conduction microphone apparatus which concerns on one Embodiment of this invention. It is a figure for demonstrating the change of the audio | voice signal when sudden noise generate | occur | produces.

  Hereinafter, a configuration example of a bone conduction microphone device according to an embodiment of the present invention will be described with reference to the drawings. In the example shown below, the case where the bone conduction microphone device is a headset used in, for example, a mobile phone will be described.

  Further, in the following example, an example will be described in which sudden noise generated due to a tooth collision sound, a jaw joint sound, or the like when a user speaks is removed. The frequency of such sudden noise is usually about 4 kHz or more (pulse width is about 0.25 msec or less). Further, when sudden noise occurs, the amplitude of the audio signal increases in an impulse shape as shown in FIG. 6B. The amplitude is A, and the amplitude when no sudden noise occurs at the same time is B. Then, the difference (A−B) between the amplitudes A and B is about twice or more the amplitude B.

[1. Configuration of bone conduction microphone device]
FIG. 1 shows a schematic configuration of a bone conduction microphone device according to an embodiment of the present invention. FIG. 1 shows only the components necessary for the above-described sudden noise removal process. 1 can be configured in the same manner as a conventional bone conduction microphone device such as a headset.

  The bone conduction microphone device 10 (microphone device) includes a bone conduction microphone 1, an analog / digital converter 2, and a signal processing module 3. The bone conduction microphone 1, the analog-digital converter 2, and the signal processing module 3 are electrically connected in this order from the sound detection side.

  The bone conduction microphone 1 can be the same as the bone conduction microphone used in a conventional bone conduction microphone device such as a headset. The bone conduction microphone 1 is used by being inserted into a user's ear, for example, specifically into the ear canal. The bone conduction microphone 1 picks up vibrations of the vocal cords transmitted to the bone when the user utters, so-called “bone conduction sound”, from the ear canal, converts it into an audio signal, and outputs the audio signal. .

  The analog-digital converter 2 has an input terminal connected to the bone conduction microphone 1 and converts an analog voice signal output from the bone conduction microphone 1 into a digital voice signal.

  The signal processing module 3 includes a sudden sound extraction unit 4 that extracts a sudden noise signal component from the voice signal, and a sudden sound removal unit 5 that removes the sudden noise signal component from the voice signal. The signal processing module 3 is not limited to the circuit unit related to the sudden noise removal processing shown in FIG. A circuit unit for performing various processing is provided.

  Here, FIG. 2 shows an internal configuration of the sudden sound extraction unit 4 and the sudden sound removal unit 5. The sudden sound extraction unit 4 includes an envelope calculation unit 11 (envelope calculation unit) and a high-pass filter 12 (filter circuit unit).

  The envelope calculation unit 11 performs peak detection on the audio signal 100 input from the analog-digital converter 2 and calculates an envelope of the audio signal 100 (hereinafter referred to as an envelope signal). The envelope calculation unit 11 can be configured by a DSP (Digital Signal Processor), for example, and can calculate an envelope signal of the audio signal 100 on software.

  FIG. 3 shows a waveform example of the envelope signal calculated by the envelope calculation unit 11. The waveform indicated by the broken line in FIG. 3 is the waveform of the audio signal 100 input from the analog-digital converter 2, and the waveform indicated by the solid line is the waveform of the envelope signal 101 calculated by the envelope calculation unit 11. In the envelope signal 101 calculated by the envelope calculation unit 11, as shown in FIG. 3, the amplitude of the envelope signal 101 increases rapidly at the time when the sudden noise 100a is generated. The envelope signal calculation method in the envelope calculation unit 11 will be described in detail later.

  The high pass filter 12 extracts a high frequency component having a predetermined frequency or higher from the envelope signal 101 input from the envelope calculation unit 11. The high pass filter 12 is configured by a digital filter such as FIR (Finite Impulse Response).

  Normally, the maximum frequency of the utterance sound detected by the bone conduction microphone 1 is about 2 to 4 kHz, and the frequency of the sudden noise generated by, for example, a tooth collision sound or a chin joint sound when speaking. Is about 4 kHz or more as described above. Therefore, when extracting such sudden noise, the high pass filter 12 is configured so that its pass band is about 4 kHz or more.

  When the envelope signal 101 calculated by the envelope calculation unit 11 is passed through the high-pass filter 12 having the above configuration, the high-pass filter 12 mainly generates a sudden noise 100a as shown in the signal waveform shown at the upper right in FIG. Only at this time, a signal 102 whose amplitude increases in a pulse shape (a signal corresponding to a signal component of sudden noise: hereinafter referred to as an extracted signal) is output.

  The sudden sound removal unit 5 includes a delay circuit 13 and an integrator 14.

  The delay circuit 13 has an input terminal connected to the output terminal of the analog-digital converter 2 and delays the audio signal 100 input from the analog-digital converter 2 for a predetermined time. Note that in the delay circuit 13, the audio corresponding to the signal delay amount in the high-pass filter 12 is set in order to align the phases of the output signal (audio signal) of the delay circuit 13 and the output signal (extracted signal) of the high-pass filter 12. Delay signal 100.

  The integrator 14 has its two input terminals connected to the output terminals of the delay circuit 13 and the high-pass filter 12, respectively, and the audio signal 100 input from the delay circuit 13 and the sudden noise input from the high-pass filter 12. The extracted signal 102 is integrated.

  At this time, however, the reciprocal of the sudden noise extraction signal 102 is integrated with the audio signal 100. That is, in the present embodiment, the integrator 14 substantially divides the audio signal 100 by the sudden noise extraction signal 102. In this case, at the time when sudden noise occurs, the inverse 1 / a of the amplitude a of the sudden noise component 102a of the extracted signal 102 is integrated into the audio signal 100, and the amplitude of the audio signal 100 is reduced (suppressed). As a result, the accumulator 14 outputs an audio signal from which sudden noise has been removed.

  In the present embodiment, an amplifier capable of gain control may be used in place of the integrator 14 so that the gain decreases as the amplitude of the sudden noise extraction signal 102 increases.

[2. Envelope signal calculation method]
Next, an example of an envelope signal calculation method in the envelope calculation unit 11 will be described. As a method for calculating the envelope signal, any method can be used as long as it can perform peak detection on the audio signal. Here, a method for calculating an envelope signal using the Hilbert transform and a method for calculating an envelope signal by comparing three amplitude values (hereinafter referred to as sampling values) of an audio signal at three consecutive sampling times are described. To do.

(1) Method Using Hilbert Transform In this method, first, the audio signal S (t) input to the envelope calculator 11 is subjected to Hilbert transform, and from the real part signal Re (t) and the imaginary part signal Im (t). An analysis signal (= Re (t) + jIm (t)) is generated. Next, the real part signal Re (t) and the imaginary part signal Im (t) are squared, respectively.

Next, the sum (Re (t) ² + Im (t) ² ) of the squared real part signal Re (t) and imaginary part signal Im (t) is calculated. Then, the envelope signal Em (t) (= [Re (t) ² + Im (t) ² ] ^1/2 ) is calculated by calculating the square root of the calculated sum signal.

(2) Method Using Level Comparison of Sampling Values FIG. 4 shows an outline of the principle when calculating the envelope signal Em (t) by comparing three sampling values of audio signals at three consecutive sampling times. In this method, first, the absolute value signal 100b (| S (t) |) of the audio signal S (t) input to the envelope calculation unit 11 is calculated.

Next, the three sampling values of the absolute value signal 100b at three consecutive sampling times are compared in magnitude. Specifically, as shown in FIG. 4, the sampling value | A _n | at a predetermined sampling time t _n , the sampling value | A _n-1 | at the time t _n−1 one sample before, and one sample Compare with the sampling value | A _{n + 1} | at a later time t _{n + 1} .

At this time, as shown in FIG. 4, when the condition of | A _n-1 | <| A _n | and | A _{n + 1} | <| A _n | is satisfied, the envelope signal Em (t _n ) at the sampling time t _n is satisfied. = | A _n | On the other hand, if the above condition is not satisfied between the three sampling values | A _n−1 |, | A _n |, and | A _{n + 1} |, the envelope signal Em (t _n ) at the sampling time t _n is sampled. The envelope signal Em (t _n-1 ) calculated at time t _n-1 is assumed (Em (t _n ) = Em (t _n-1 )). By repeating this calculation step at each sampling time, the envelope signal Em (t) of the audio signal S (t) can be calculated.

  Here, comparing the two calculation methods (1) and (2) described above, the calculation method (1) is more advantageous from the viewpoint of accuracy, but the processing amount (processing time), downsizing of the apparatus, and the like From the viewpoint, the calculation method (2) is more advantageous. Therefore, the calculation method (1) is used in applications where the calculation accuracy of the envelope signal Em (t) is required, and the calculation method (2) is used in applications where high-speed processing and / or downsizing is required. preferable.

  When the envelope calculation unit 11 is configured by a DSP as in the present embodiment, the above-described calculation process of the envelope signal Em (t) can be realized on software, but the present invention is not limited to this. Not. You may comprise the processing circuit of each calculation process of the envelope signal Em (t) mentioned above with hardware.

[3. Operation of bone conduction microphone device]
Next, an operation example of the bone conduction microphone device 10 of this embodiment will be described with reference to FIGS. 5 (a) and 5 (b). 5A is a waveform diagram of a signal output from each part of the bone conduction microphone device 10, and FIG. 5B is a schematic configuration diagram of the bone conduction microphone device 10.

  First, the bone conduction microphone 1 mounted in one ear of the user detects an audio signal (analog signal) when the user speaks and outputs the audio signal to the analog-to-digital converter 2. Next, the analog-digital converter 2 converts the input analog audio signal into a digital audio signal, and outputs the digital audio signal to the sudden sound extraction unit 4 and the sudden sound removal unit 5 in the signal processing module 3. To do.

Here, the left diagram in FIG. 5A shows the waveform of the audio signal 200 input from the analog-digital converter 2 to the sudden sound extraction unit 4. In this example, sudden noise 2 occurs twice in the audio signal 200.
The case where 00a occurs is shown.

  Next, the envelope calculation unit 11 in the sudden sound extraction unit 4 calculates the envelope signal from the input audio signal 200 using, for example, the calculation methods (1) and (2) described above, and the envelope The signal is output to the high pass filter 12. Next, the high-pass filter 12 has a high frequency component higher than a predetermined frequency (for example, about 4 kHz) included in the input envelope signal, that is, a component of the sudden noise 200a (abrupt noise component 202a in the center diagram of FIG. 5A). To extract. Then, the high-pass filter 12 outputs the generated sudden noise extraction signal to the sudden sound removal unit 5.

  Here, the central diagram in FIG. 5A shows the waveform of the sudden noise extraction signal 202 input from the sudden sound extraction unit 4 to the sudden sound removal unit 5. In this example, the extraction signal 202 whose amplitude has increased rapidly in a pulse shape is output from the sudden sound extraction unit 4 at two generation times of the sudden noise 200a.

  Next, the sudden sound removal unit 5 is based on the audio signal 200 delayed by a predetermined time by the delay circuit 13 provided therein and the sudden noise extraction signal 202 input from the sudden sound extraction unit 4. The sudden noise component is removed from the audio signal 200. In the present embodiment, the reciprocal of the sudden noise extraction signal 202 is added to the audio signal 200. As a result, the amplitude of the audio signal 200 is suppressed at two occurrence times of the sudden noise 200a, and the sudden noise component can be removed.

  Here, the right diagram in FIG. 5A shows the waveform of the audio signal output from the sudden sound removal unit 5. In the audio signal output from the sudden sound removal unit 5, the amplitude of the two occurrence times of the sudden noise 200a is suppressed, and the sudden noise 200a is removed. In the bone conduction microphone device 10 of the present embodiment, the sudden noise 200a is removed from the audio signal 200 detected by the bone conduction microphone 1 as described above.

  As described above, the bone conduction microphone device 10 according to the present embodiment has a simple structure and can easily remove the influence of sudden noise generated in the body from an audio signal. Therefore, according to the present embodiment, a bone conduction microphone device having a simple structure and capable of a higher quality and more comfortable call can be provided.

  In the bone conduction microphone device 10 of the above embodiment, the example in which the sudden sound extraction unit 4 is configured by the envelope calculation unit 11 and the high-pass filter 12 has been described, but the present invention is not limited to this. For example, when the envelope calculation unit 11 uses the Hilbert transform to calculate an envelope signal (envelope) of an audio signal, when the processing frequency is set to a predetermined frequency or higher (for example, 4 kHz or higher), the high-pass filter 12 May not be used.

  More specifically, when the processing frequency when the envelope calculation unit 11 performs the Hilbert transform on the audio signal is set to 4 kHz or more, for example, an analysis signal corresponding to the signal component of the sudden noise is obtained by the Hilbert transform. In this case, the envelope calculation unit 11 calculates the envelope signal using the real part and the imaginary part of the analysis signal, thereby directly obtaining the extraction signal (extraction signal 102 in FIG. 2) corresponding to the signal component of the sudden noise. Can be obtained.

  That is, in this case, the envelope calculation unit 11 can have both an envelope signal calculation function and a sudden noise signal component extraction function (high-pass filter function). Therefore, in this case, the high-pass filter 12 shown in FIG. 2 is not necessary, and the configuration of the sudden sound extraction unit 4 can be simplified.

  Moreover, in the said embodiment, although the headset used, for example with a mobile telephone etc. was mentioned as an example as a bone-conduction microphone apparatus, this invention is not limited to this. In applications where a bone conduction microphone is used and internal noise having a frequency higher than the speech sound is detected by the bone conduction microphone, the present invention can be applied to any device, and the same effect can be obtained.

  DESCRIPTION OF SYMBOLS 1 ... Bone conduction microphone, 2 ... Analog-digital converter, 3 ... Signal processing module, 4 ... Sudden sound extraction part, 5 ... Sudden sound removal part, 10 ... Bone conduction microphone apparatus, 11 ... Envelope calculation part, 12 ... High pass filter , 13 ... delay circuit, 14 ... integrator, 100, 200 ... audio signal, 100a, 200a ... sudden noise, 101 ... envelope signal, 102, 202 ... extracted signal, 102a, 202a ... sudden noise component

Claims (4)

With bone conduction microphones,
A sudden sound extraction unit that calculates an envelope of an audio signal detected by the bone conduction microphone and extracts a signal corresponding to a signal component of sudden noise included in the audio signal based on the envelope;
A sudden sound removing unit that removes the sudden noise signal component from the audio signal based on a signal corresponding to the sudden noise signal component extracted by the sudden sound extraction unit ;
The sudden sound extraction unit
An envelope calculation unit for calculating an envelope of the audio signal;
A microphone device having a filter circuit unit that extracts a signal having a predetermined frequency or more from the envelope and outputs the extracted signal as a signal corresponding to the signal component of the sudden noise . The envelope calculation section compares the three high and low amplitude values of the audio signal at the three sampling successive times, the microphone apparatus according to claim 1 for calculating the envelope of the audio signal. The envelope calculation section, the speech signal Hilbert transform to calculate the analytic signal of the audio signal, to claim 1 for calculating the envelope of the audio signal based on the real part and the imaginary part of the analytic signal The microphone device described. With bone conduction microphones,
A sudden sound extraction unit that calculates an envelope of an audio signal detected by the bone conduction microphone and extracts a signal corresponding to a signal component of sudden noise included in the audio signal based on the envelope;
A sudden sound removing unit that removes the sudden noise signal component from the audio signal based on a signal corresponding to the sudden noise signal component extracted by the sudden sound extraction unit;
The sudden sound extraction unit
Calculating an analysis signal of the audio signal by performing a Hilbert transform on the audio signal at a processing frequency equal to or higher than a predetermined frequency, calculating an envelope of the audio signal based on a real part and an imaginary part of the analysis signal; and An envelope calculation unit that outputs an envelope as a signal corresponding to the signal component of the sudden noise;
Microphone device.

Images