JP4568905B2 - Microphone device and speech detection device - Google Patents

Description

The present invention relates to a microphone apparatus and a speech detection apparatus.

  In recent years, microphone devices, commonly known as “microphones”, have expanded beyond the field of video and broadcasting to a variety of worlds. In response to this widespread use, the times demand increasingly smaller and higher performance microphones. For example, in a navigation system, accurate information transmission between an instruction sending side and a receiving side is important.

  Also, in the field of speech recognition, which has been actively researched in recent years, the microphone device plays a major role. For example, when a microphone device picks up sound (so-called noise) from a person other than the speaker, it often causes a decrease in recognition (deterioration of recognition). In order to prevent such degradation of recognition caused by noise, a microphone device capable of performing voice transmission after removing noise has been considered (for example, Patent Document 1).

When performing recognition rate improvement processing such as noise removal, it is necessary to acquire information on the time (voice segment) during which the speaker is speaking and perform processing in accordance with the speech. For example, if the information of the voice section is acquired by mistake, the processing before the speaker starts speaking will be confused with the state where the speaker has already started speaking, and the recognition rate will improve as expected. Not. Therefore, as a method for appropriately acquiring the information of the voice section, research for detecting the movement of the human lips by using image recognition, a camera or the like has been performed at ATR (International Telecommunications Research Institute). Yes.
JP 2003-111186 A

  As described above, when a method for detecting the movement of a human lip by image recognition or a camera is used, the equipment itself for realizing the method becomes very large and the amount of money becomes expensive. Also, CPU resources for realizing the entire system become large.

  The present invention has been made in view of the above circumstances, and an object thereof is to easily realize a microphone device that can efficiently detect a speaker's speech period.

In order to achieve the above object, a microphone device according to a first aspect of the present invention includes:
Ultrasonic transmission means for transmitting ultrasonic waves to be applied to the speaker;
Sound wave receiving means for receiving sound waves from the speaker's utterance;
Wherein arranged in the proximity of the ultrasonic delivery means, the reflected wave reception ultrasonic wave transmitted from the ultrasonic transmitting means receives the reflected wave reflected when the speaker while being constituted by a common receiving means and the wave reception unit Means,
Utterance discriminating means for discriminating whether or not a speaker is speaking based on the reflected wave received by the reflected wave receiving means ;
Only when it is determined that the speaking speaker by the speech discriminating means, and output means for outputting sound based on the sound wave received by the wave receiving means,
It is characterized by that.

The receiving means receives a mixed wave of the sound wave and the reflected wave, and converts the mixed wave into an electric signal,
Separating means for separating the electric signal of the mixed wave into a signal corresponding to a sound wave and a signal corresponding to a reflected wave,
The utterance discrimination means preferably discriminates whether or not the speaker is speaking based on the separated reflected wave signal.

  It is preferable that the separation unit separates the mixed wave into a sound wave and an ultrasonic wave based on a frequency of a component included in the mixed wave signal.

The utterance discrimination means is
Detecting means for detecting a reflected wave received by the reflected wave receiving means and extracting a signal waveform; and
On the basis of the signal waveform extracted by the detection means, speaker lips and a speech activity detector for determining whether or not the motion indicating the speech operation,
It is preferable.

  The speech action detecting means detects a time change of the amplitude value of the signal waveform extracted by the detection means, determines whether the time change is larger than a predetermined threshold, and the speaker's lips are made to speak. It is preferable to determine whether or not it is moving.

  It is preferable that the ultrasonic transmission means and the reception means are constituted by a common dynamic microphone.

In order to achieve the above object, an utterance detection apparatus according to the second aspect of the present invention provides:
Ultrasonic transmission means for transmitting ultrasonic waves to be applied to the speaker;
Sound wave receiving means for receiving sound waves from the speaker's utterance;
Wherein arranged in the proximity of the ultrasonic delivery means, the reflected wave reception ultrasonic wave transmitted from the ultrasonic transmitting means receives the reflected wave reflected when the speaker while being constituted by a common receiving means and the wave reception unit Means,
On the basis of the reflected wave received by the reflection wave receiving means, and a speech discrimination means for discriminating whether the speaking speaker,
It is characterized by that.

The receiving means receives a mixed wave of the sound wave and the reflected wave, and converts the mixed wave into an electric signal,
Separating means for separating the electric signal of the mixed wave into a signal corresponding to a sound wave and a signal corresponding to a reflected wave,
The utterance discrimination means preferably discriminates whether or not the speaker is speaking based on the separated reflected wave signal.

The utterance discrimination means is
Detecting means for detecting a reflected wave received by the reflected wave receiving means and extracting a signal waveform; and
On the basis of the signal waveform extracted by the detection means, speaker lips and a speech activity detector for determining whether or not the motion indicating the speech operation,
It is preferable.

  ADVANTAGE OF THE INVENTION According to this invention, the microphone apparatus which can detect a speaker's speech period is realizable.

  As shown in FIG. 1, the microphone device 1 according to the embodiment of the present invention includes a microphone unit 11, a band separation unit 12, a detection circuit 13, a differentiation circuit 14, an utterance determination unit 15, and a speaker switch 16. And comprising. The microphone unit 11 includes an ultrasonic transmission unit 111 and a sonic ultrasonic reception unit 112. The band separation unit 12 includes an LPF (Low Pass Filter) 121 and a BPF (Band Pass Filter) 122. Further, the band separation unit 12, the detection circuit 13, the differentiation circuit 14, the utterance determination unit 15, and the speaker switch 16 are stored inside the housing K.

The microphone unit 11 is a portion that is usually used as an input port for spoken speech that is located near the mouth of a speaker (hereinafter referred to as speaker S) who wants to transmit speech using the microphone device 1.
The ultrasonic transmission unit 111 is composed of a small ultrasonic oscillator such as a transducer, and emits ultrasonic waves of a predetermined frequency (for example, about 40 kHz) as the power supply (not shown) of the microphone device 1 is turned on. Continuously send to a predetermined external range.
The sonic wave ultrasonic reception unit 112 is configured by an ECM (electret, condenser, microphone) or the like, and collects sound (sound wave) emitted by the speaker S. The ultrasonic ultrasonic wave reception unit 112 receives the reflected wave that bounces off the ultrasonic speaker S emitted from the ultrasonic wave transmission unit 111. When the sound wave and the ultrasonic wave are received, the sound wave ultrasonic wave reception unit 112 converts them into an electric signal and supplies it to the band separation unit 12.

  The band separation unit 12 separates the electrical signal supplied from the sonic ultrasonic wave reception unit 112 into a signal corresponding to the sound wave and a signal corresponding to the ultrasonic wave based on the frequency. More specifically, the LPF 121 passes sound waves, and the BPF 122 passes ultrasonic waves. That is, the LPF 121 passes only a signal having a frequency equal to or lower than a predetermined cutoff frequency (about 20 kHz) and attenuates a signal having a frequency equal to or higher than the cutoff frequency. On the other hand, the BPF 122 passes only signals having a frequency in a predetermined frequency range (around 40 kHz) and attenuates signals having other frequencies.

  The detection circuit 13 is a circuit that detects the signal waveform supplied from the BPF 122, and includes a diode for half-wave rectification and a capacitor for smoothing the signal waveform.

  The differentiating circuit 14 is a circuit that outputs a signal corresponding to a signal waveform differentiated with respect to time with respect to the signal waveform detected by the detection circuit 13 (a signal corresponding to a temporal change in amplitude value). Here, the signal that has passed through the detection circuit 13 is a reflected wave that bounces off the speaker S, and the signal corresponding to the time change is a signal that indicates the operation of the speaker S (specifically, the movement of the lips). . By passing through the differentiating circuit 14, the signal waveform indicating the movement of the body of the speaker S, that is, the slow movement of the face to the front, back, left and right is not extracted, but only the signal waveform indicating the movement of the lips is extracted.

  The utterance determination unit 15 includes a comparator or the like as shown in the figure, and determines whether or not the signal supplied from the differentiating circuit 14 and corresponding to the time change of the amplitude value exceeds a predetermined threshold value. That is, the speed of movement of the speaker's lips is detected, and whether or not the movement of the lips of the speaker S is at a level indicating utterance depending on whether or not the speed exceeds a predetermined threshold value. Determine. When it is determined that the threshold value is exceeded, a signal of a predetermined level that requests the switch on for audio output to be turned on is output to the speaker switch 16.

  The speaker switch 16 is composed of a momentary switch, and when a signal of a predetermined level is supplied from the utterance determination unit 15, the switch for sound output is set to ON. If the speaker switch 16 does not receive a signal of a predetermined level from the utterance determination unit 15, the speaker output 16 continues to maintain the voice output in an OFF state (not switched on). More specifically, since the differentiation circuit 14 outputs a signal indicating that the lips have moved greatly and quickly as a signal, for example, when the mouth is slowly opened wide or when it does not move as it is moved, the change is differentiated. The circuit 14 is cut and the speaker switch 16 is not turned on.

  Next, processing operations performed by the microphone device 1 according to the present embodiment will be described with reference to the flowchart of FIG. Here, it is assumed that the speaker S holds the microphone device 1 in his hand, grasps the housing K, and brings the microphone unit 11 close to the mouth as shown in FIG.

  The ultrasonic transmission unit 111 in the microphone unit 11 continuously transmits ultrasonic waves having a predetermined frequency to the outside (step S101). A part of the ultrasonic wave hits the lip of the speaker S and generates a reflected wave.

  The sonic ultrasonic wave reception unit 112 receives the reflected wave generated by hitting the speaker S by the transmission of the ultrasonic wave in step S101 (step S102). At that time, if the speaker S is uttering a word, the voice (sound wave) is also received at the same time.

  The sonic wave ultrasonic wave reception unit 112 converts the mixed wave including the received sound wave and the ultrasonic wave into an electric signal and supplies the electric signal to the band separation unit 12. In the band separation unit 12, the LPF 121 and the BPF 122 perform band separation of the mixed wave (step S103). Here, the LPF 121 passes a signal having a frequency of 20 kHz or less as a sound wave. On the other hand, the BPF 122 passes a signal having a frequency of around 40 kHz as equivalent to an ultrasonic wave. The ultrasonic wave that has passed through the BPF 122 is output toward the detection circuit 13.

  The detection circuit 13 detects the signal supplied through the BPF 122, performs half-wave rectification and waveform smoothing (step S104), and supplies the signal to the differentiation circuit 14.

  The differentiating circuit 14 performs processing corresponding to time differentiation (processing for detecting temporal changes in the waveform) on the waveform detected by the detection circuit 13 (step S105). Then, a signal waveform corresponding to the time change is supplied to the utterance determination unit 15.

  The utterance determination unit 15 determines whether or not the signal level of the signal corresponding to the time change output from the differentiation circuit 14 is greater than a predetermined threshold (step S106).

  If it is determined that the value is larger than the threshold value (step S106: Yes), the speech determination unit 15 determines that the lips of the speaker S are greatly opened quickly, that is, starts to speak, and supplies a signal of a predetermined level to the speaker switch 16. The voice output is switched to ON (step S107). On the other hand, if it is determined that the time change of the waveform is equal to or less than the threshold (step S106: No), the utterance determination unit 15 does not supply a signal to the speaker switch 16, and does not switch ON / OFF the audio output. Let it be maintained.

  Through the above processing operation, the microphone device 1 automatically switches the audio output switch (speaker switch 16). That is, when it is determined that the movement of the lips of the speaker S is greater than a predetermined size and speed, the speaker S is considered to be speaking and the voice output switch is turned on.

  As described above, according to the microphone device 1 according to the present embodiment, it is possible to detect the movement of the speaker's lips by ultrasonic waves, without a large accessory device such as an image recognition device or a camera. The speaker's utterance period can be detected.

In addition, this invention is not limited to the example shown by the said embodiment, A various deformation | transformation and application are possible.
For example, in the above embodiment, the microphone unit 11 is configured by using the ultrasonic transmission unit 111 and the ultrasonic ultrasonic reception unit 112. However, the present invention is not limited to this, and the microphone device 1 can be configured using, for example, a dynamic microphone. In this case, the diaphragm constituting the dynamic microphone plays a role of both the ultrasonic ultrasonic receiving unit 112 and the ultrasonic transmission unit 111.

  Moreover, in the said embodiment, although the sonic wave ultrasonic wave reception part 112 shall perform both reception of a sonic wave and reception of an ultrasonic wave, the sonic wave reception part which receives a sonic wave, and the ultrasonic wave which receives an ultrasonic wave It is good also as a structure which provides a receiving part separately. However, by performing reception of sound waves and reception of ultrasonic waves simultaneously from the same portion (the sound wave ultrasonic wave reception unit 112) as in the above embodiment, the speaker is directed to receive sound waves and receive ultrasonic waves. You can speak without considering gender differences. More specifically, without considering the directivity difference between the ultrasonic wave receiving unit and the sound wave receiving unit as shown in FIG. 4 (a), the speaker S can be connected to the receiving unit as shown in FIG. 4 (b). You can speak in front of you.

  In the above embodiment, the microphone device 1 is described as a hand microphone that the speaker S holds and holds the housing K. However, the form of the microphone device 1 is not limited to this, and the speaker S is not limited thereto. However, it can be realized as a pin microphone to be worn on its chest or a hanging microphone that picks up sound from above. Moreover, you may install in the dashboard etc. of a motor vehicle toward a driver | operator's lip. At that time, in order to match the reception range of sound waves, it is necessary to limit the transmission direction of ultrasonic waves to a narrow range for pin microphones and microphones installed on dashboards. Must be set.

  In the above embodiment, the threshold value is set to determine the magnitude of the time change of the waveform. However, this threshold value is not limited to a specific value. This is because, for example, children and adults, Westerners and Japanese, etc. have different ways of opening their mouths when speaking, and by setting a threshold value for each category, it is more accurate. It becomes possible to detect the utterance period of the speaker.

  Further, the detailed circuit configuration such as the detection circuit and the differentiation circuit is not limited to the example shown in the above embodiment, and it is sufficient that each circuit can be configured to achieve each purpose.

1 is a configuration diagram of a microphone device according to an embodiment of the present invention. 3 is a flowchart for explaining a processing operation performed by the microphone device of FIG. 1. It is a figure which shows typically a mode when the microphone apparatus of FIG. 1 is utilized by the speaker. It is a schematic diagram for demonstrating that an utterance can be performed without considering the difference in directivity between a sound wave and an ultrasonic wave when the sound wave receiving unit and the ultrasonic wave receiving unit are the same.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Microphone apparatus 11 Microphone unit 111 Ultrasonic transmission part 112 Sonic wave ultrasonic reception part 12 Band separation part 121 LPF
122 BPF
13 Detection Circuit 14 Differentiation Circuit 15 Utterance Determination Unit 16 Speaker Switch

Claims (9)

Ultrasonic transmission means for transmitting ultrasonic waves to be applied to the speaker;
Sound wave receiving means for receiving sound waves from the speaker's utterance;
Wherein arranged in the proximity of the ultrasonic delivery means, the reflected wave reception ultrasonic wave transmitted from the ultrasonic transmitting means receives the reflected wave reflected when the speaker while being constituted by a common receiving means and the wave reception unit Means,
Utterance discriminating means for discriminating whether or not a speaker is speaking based on the reflected wave received by the reflected wave receiving means ;
Only when it is determined that the speaking speaker by the speech discriminating means, and output means for outputting sound based on the sound wave received by the wave receiving means,
A microphone device characterized by that. The receiving means receives a mixed wave of the sound wave and the reflected wave, and converts the mixed wave into an electric signal,
Separating means for separating the electric signal of the mixed wave into a signal corresponding to a sound wave and a signal corresponding to a reflected wave,
The speech discrimination means determines whether or not the speaker is speaking based on the separated reflected wave signal.
The microphone device according to claim 1. The separation means separates the mixed wave into a sound wave and an ultrasonic wave based on a frequency of a component included in the signal of the mixed wave;
The microphone device according to claim 2. The utterance discrimination means is
Detecting means for detecting a reflected wave received by the reflected wave receiving means and extracting a signal waveform; and
On the basis of the signal waveform extracted by the detection means, speaker lips and a speech activity detector for determining whether or not the motion indicating the speech operation,
The microphone device according to claim 1, 2, or 3. The speech action detecting means detects a time change of the amplitude value of the signal waveform extracted by the detection means, determines whether or not the time change is larger than a predetermined threshold, and the speaker's lips make a speech action To determine whether or not it is moving
The microphone device according to claim 4. The ultrasonic transmission means and the reception means are composed of a common dynamic microphone.
The microphone device according to any one of claims 1 to 5, wherein Ultrasonic transmission means for transmitting ultrasonic waves to be applied to the speaker;
Sound wave receiving means for receiving sound waves from the speaker's utterance;
Wherein arranged in the proximity of the ultrasonic delivery means, the reflected wave reception ultrasonic wave transmitted from the ultrasonic transmitting means receives the reflected wave reflected when the speaker while being constituted by a common receiving means and the wave reception unit Means,
On the basis of the reflected wave received by the reflection wave receiving means, and a speech discrimination means for discriminating whether the speaking speaker,
An utterance detection device characterized by that. The receiving means receives a mixed wave of the sound wave and the reflected wave, and converts the mixed wave into an electric signal,
Separating means for separating the electric signal of the mixed wave into a signal corresponding to a sound wave and a signal corresponding to a reflected wave,
The speech discrimination means determines whether or not the speaker is speaking based on the separated reflected wave signal.
The utterance detection device according to claim 7. The utterance discrimination means is
Detecting means for detecting a reflected wave received by the reflected wave receiving means and extracting a signal waveform; and
On the basis of the signal waveform extracted by the detection means, speaker lips and a speech activity detector for determining whether or not the motion indicating the speech operation,
The utterance detection apparatus according to claim 7 or 8, wherein

Images