JP2008042740A - Non-audible murmur pickup microphone - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a non-audible murmur pickup microphone (NAM microphone) capable of picking up a non-audible voice, and converting it into a voice signal that a listener and a voice recognition means easily recognize and then outputting the voice signal with simple constitution. <P>SOLUTION: The NAM microphone X includes an NAM propagation unit 12 as a soft member which is brought into contact with a skin surface 1a to propagate a non-audible murmur (NAM) propagated in the human body, a microphone 11 which transduces the NAM propagated in the NAM propagation unit 12 into an electric signal, and a high-pass filter 22 which performs high-pass filter processing of about 1000 Hz in cutoff frequency for a signal of a voice obtained by the microphone 11. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention relates to a non-audible tweet voice collecting microphone for collecting a non-audible tweet voice.

Recently, with the spread of mobile phones and their communication networks, it is possible to communicate by voice (conversation) with other people anytime and anywhere. Furthermore, various devices such as a personal computer and a car navigation device equipped with voice recognition means can execute processing in response to a voice command.
On the other hand, there are many situations where utterances are restricted to prevent inconvenience to surrounding people, such as in trains and libraries, and utterances are restricted because the contents of conversations are confidential matters. Even in situations where utterances are restricted in this way, if voice commands and voice commands to devices can be issued without leaking the content of utterances to the surroundings, voice communication can be further on-demand, or voice devices can be remotely controlled. Control etc. is promoted and it leads to the improvement of efficiency of various operations.
Further, even a disabled person who cannot speak normal speech due to a disorder in the pharynx (such as vocal cords) can often speak with a non-audible muttering voice. For this reason, if it becomes possible to make a call with a non-audible muttering voice or give a command to the device, the convenience of the handicapped person with such a throat is greatly improved.
On the other hand, Patent Document 1 proposes a communication interface system that inputs voice by collecting non-audible murmur voice (NAM: Non-Audible Murmur). A non-audible murmur voice (NAM) is a voice (unvoiced sound) that does not accompany regular vibration of the vocal cords, and is a vibration sound (breathing sound) that propagates through a non-audible soft tissue in the body from the outside. In other words, the sound is a breathing sound that is not accompanied by vocal cord vibration that occurs in the human vocal tract. For example, in a soundproof room environment, a non-audible voice (breathing sound) that cannot be heard by people around 1 to 2 meters away is defined as a “non-audible muttering voice” and the vocal tract (especially the oral cavity) is narrowed down. An audible voice that produces an unvoiced sound to the extent that it can be heard by people around 1 to 2 meters away by increasing the flow velocity of the air passing through the road is defined as an “audible whispering voice”.
Such a non-audible murmur voice signal cannot be collected by a normal microphone that detects vibration in the acoustic space, and is therefore generally collected by a meat conduction microphone that collects body conduction sound in the body. Conventionally, this meat conduction microphone is mainly used for collecting non-audible whispering sound (NAM), so it is also called a NAM microphone, and details thereof are disclosed in Patent Document 1 and the like.
This NAM microphone (meat conduction microphone) converts a non-audible murmur sound propagating through a soft member made of silicon or the like that propagates a non-audible murmur sound by being in close contact with the skin surface of a human body into an electrical signal. And a microphone. The NAM microphone is attached so that the soft member is in close contact with the skin surface on the thoracic papillary muscle, just below the mastoid process of the skull in the lower part of the auricle, and the soft composition ( Meat conduction sounds (non-audible whispering sounds) transmitted through muscles and fat other than bone) are collected.
Here, since the NAM microphone mainly collects only the sound that propagates through the soft composition of the human body, even if the surrounding noise (noise sound) is large, the human body functions as a noise removal filter. As a result, an acoustic signal having a high S / N ratio can be collected. That is, the NAM microphone has high noise resistance against sound propagating in the air.

By the way, as long as the audible whispering sound is amplified to a sufficient volume, even a general person who has not received special training can hear the utterance content at a high recognition rate.
One problem is that the inaudible tweet voice has a problem that even if the voice is simply amplified, it is difficult for the listener to hear the utterance content (recognition rate is low). The same applies to the voice recognition means.
On the other hand, for example, in Non-Patent Document 1, a signal of a non-audible muttering voice obtained by a NAM microphone (meat conduction microphone) based on a mixed normal distribution model which is an example of a model based on a statistical spectrum conversion method is usually used. A technique for converting a voice signal (voiced sound) into a signal is shown.
Further, Patent Document 2 estimates the pitch frequency of a normal uttered sound (voiced sound) by comparing the power of inaudible murmur voice signals obtained by two NAM microphones (meat conduction microphones), and the estimation result. Based on the above, a technique for converting a signal of a non-audible muttering voice into a signal of voice (voiced sound) normally uttered is shown.
By using the techniques shown in Non-Patent Document 1 and Patent Document 1, a non-audible muttering voice signal obtained through a body conduction microphone is converted into a normal voice (voiced sound) signal that is relatively easy for the listener to hear. Can be converted.
WO2004 / 021738 pamphlet JP 2006-086877 A Toda Tomomoto et al. “Conversion from non-audible murmur (NAM) to normal speech based on mixed normal distribution model” IEICE Technical Report, SP2004-107, pp.67-72, December 2004

However, the conventional technique (the technique disclosed in Patent Document 1 and Patent Document 2) that converts a non-audible murmur sound into a normal audible sound based on the sound conversion model has a delay in sound signal transmission by the time required for the conversion. It had the problem that it occurred. Furthermore, since the apparatus that realizes the conventional technique requires an arithmetic means (microcomputer) that executes signal conversion processing based on a voice conversion model, the power consumption is relatively large and the cost is high. Had. Here, when performing a voice call or a voice command to a device, it is necessary to configure a portable or body-worn voice input / output device. However, in that case, the problem of high power consumption is sufficient. It also leads to the problem that continuous use time cannot be secured.
In addition, as shown in Patent Document 2, the non-audible murmur sound is an unvoiced sound that does not involve regular vibration of the vocal cords. Then, as shown in Patent Document 1 and Patent Document 2, when converting an inaudible muttering voice signal, which is an unvoiced sound, into a normal voice (voiced sound) signal, a conversion characteristic of an acoustic feature amount by the vocal tract ( A combination of a vocal tract feature value conversion model that represents the conversion characteristics of input signal feature values to output signal feature values) and a vocal cord feature value conversion model that represents acoustic feature value conversion characteristics of the sound source (voice vocal cords) A speech conversion model is used. Processing using such a speech conversion model includes processing for generating (estimating) “existing” from “absent” regarding voice pitch information. For this reason, if a non-audible tweet signal is converted into a normal (voiced) signal, a signal containing unnatural intonation or incorrect speech that is not uttered can be obtained. There was a problem that the speech recognition rate of the recognition means was lowered.
Accordingly, the present invention has been made in view of the above circumstances, and the object of the present invention is to collect a non-audible murmur voice with a very simple configuration, which can be easily recognized by the listener or voice recognition means. An object of the present invention is to provide a non-audible tweet voice collecting microphone (NAM microphone) which can be converted into a signal and output.

In order to achieve the above object, the present invention includes the components shown in the following (1) to (3), so that a non-audible murmur voice (a voice by a breathing sound without a vocal cord vibration generated in a human vocal tract). ) Is a non-audible muttering voice collecting microphone (hereinafter referred to as a NAM microphone).
(1) A soft member that propagates the inaudible murmur sound by being in close contact with the skin surface of a human body.
(2) A microphone that converts the inaudible murmur sound propagating through the soft member into an electric signal.
(3) A high-pass filter that performs a high-pass filter process on the inaudible tweet signal obtained by the microphone.
Here, the cut-off frequency of the high-pass filter is preferably about 800 to 1400 Hz, for example. In that case, the slope characteristic of the high-pass filter is preferably about −16 to −14 dB / oct.
In this way, a signal obtained by passing a non-audible murmur voice signal through the high-pass filter (the output signal of the high-pass filter) can be heard almost equivalently to the audible whisper voice signal simply by amplifying it. It turns out that it becomes an easy (recognizable) signal. Moreover, the high-pass filter can be realized by a very simple circuit that consumes very little power (for example, it can be realized by a capacitor and a resistance element).
In general, the microphone includes an amplifier that amplifies an audio signal. However, the NAM microphone according to the present invention includes an amplifier that amplifies the signal of the inaudible murmur audio between the microphone and the high-pass filter. It is desirable.
This makes it possible to amplify the inaudible murmur voice signal having a low volume level to a level sufficient as an input signal of the high-pass filter.

According to the NAM microphone of the present invention, it is possible to obtain an audio signal that is easy for the listener to hear by collecting inaudible tweets. The ease of listening is comparable to the audible whispering sound.
In addition, the sound obtained by the NAM microphone according to the present invention is converted based on a model obtained by combining a normal sound (a non-audible sound signal obtained by a conventional method) with a vocal tract feature value conversion model and a sound source feature value conversion model. The normal voice (voiced sound)) is stable without including unnatural voices that are not intonation and false voices that are not originally uttered.
Furthermore, according to the present invention, signal conversion processing (relatively high load processing) based on the voice conversion model is unnecessary, there is no delay in voice signal transmission, and arithmetic means (microcomputer) for executing the processing is provided. Is also unnecessary. For this reason, the NAM microphone according to the present invention may increase the weight and volume of the device and shorten the continuous use time even when incorporated in a small device such as a mobile phone or a body-worn device. rare.

Embodiments of the present invention will be described below with reference to the accompanying drawings for understanding of the present invention. In addition, the following embodiment is an example which actualized this invention, Comprising: It is not the thing of the character which limits the technical scope of this invention.
FIG. 1 is a schematic configuration diagram (partial block diagram) of the NAM microphone X according to the embodiment of the present invention, FIG. 2 is a schematic diagram showing a state in which the NAM microphone X is attached to a human body, and FIG. FIG. 4 is a graph showing the voiceprint of a murmured voice and the voiceprint of an audible whispering voice, FIG. 4 is a graph showing the result of a first evaluation experiment (naturalness evaluation experiment) of the NAM microphone X, and FIG. 5 is a second evaluation experiment of the NAM microphone X. It is a graph showing the result of a word recognition accuracy evaluation experiment.

First, the configuration of the NAM microphone X according to the embodiment of the present invention will be described with reference to FIG.
Here, FIG. 1A is an overall configuration diagram of the NAM microphone X, and FIG. 1B is a front view of the sound detection unit 10 constituting a part of the NAM microphone X (viewed from the skin contact surface side). FIG. 1C is a block diagram showing the configuration of the signal processing unit 20 that constitutes a part of the NAM microphone X. In addition, in Fig.1 (a), about the audio | voice detection part 10, sectional drawing is shown.
The NAM microphone X is a non-audible murmur voice collecting microphone that collects a non-audible murmur voice (NAM), which is a voice generated by a breathing sound that is not accompanied by vocal cord vibration generated in the human vocal tract.
As shown in FIG. 1A, the NAM microphone X includes an audio detection unit 10, a signal processing unit 20, and a signal line 30 that connects them, and the signal processing unit 20 includes an ear mounting casing 27. Is housed in.
The sound detection unit 10 is attached to the skin surface of the human body, thereby generating vibrations of non-audible muttering sound (NAM) generated by the vocal tract of the human body and propagating through the soft tissue (mainly the meat part) in the body as an electrical signal. To convert.
The signal processing unit 20 performs various processing (signal amplification processing, transmission processing to an external device, etc.) on the detection signal (audio signal) by the audio detection unit 10.

Hereinafter, the voice detection unit 10 will be described with reference to FIGS. 1 (a) and 1 (b).
The voice detection unit 10 includes a microphone 11, a NAM propagation unit 12, an inner cover member 13, an adhesive sound insulation unit 14, and an outer cover member 15.
The NAM propagation unit 12 is a soft member that propagates a non-audible murmur sound propagating through the human body when its one surface 12 a is in close contact with the skin surface 1 a of the human body 1. The NAM propagation part 12 is made of a material whose acoustic impedance characteristic is close to the acoustic impedance characteristic of the flesh in the human body, such as urethane elastomer or silicon. Thereby, a non-audible murmur voice can be efficiently propagated from the human body (skin) to the NAM propagation unit 12.
The microphone 11 converts an inaudible murmur sound (vibration) propagating through the NAM propagation unit 12 (soft member) into an electric signal. The microphone 11 is in direct contact with the NAM propagation unit 12 as a whole of the sound sensing unit (the left side surface in FIG. 1A) that senses vibrations of sound. Thereby, the microphone 11 detects the vibration (sound) of the NAM propagation unit 12 with high sensitivity. The output signal of the microphone 11 (inaudible tweet voice signal) is transmitted to the signal processing unit 20 through the signal line 30.
The inner cover member 13 covers the entire portion other than the contact surface (hereinafter referred to as the skin contact surface 12a) with the skin surface 1a of the NAM propagation portion 12. That is, the inner cover member 13 is a container-like member whose one surface is in an open state. The microphone 11 is housed inside the inner cover member 13 and the NAM propagation portion 12 (soft member) is filled with almost no gap. (The microphone 11 is embedded in the NAM propagation unit 12).
The adhesive sound insulation part 14 is made of urethane elastomer, which is an adhesive soft member, and is formed so as to cover the entire outside of the inner cover member 13. Furthermore, the adhesive sound insulating part 14 has a part (hereinafter referred to as a skin adhesive part 14 a) that contacts and adheres to the skin surface 1 a of the human body 1. The skin adhesive portion 14a is formed over the entire contact surface (hereinafter referred to as the skin contact surface 12a) with the skin 1a in the NAM propagation portion 12.
The outer cover member 15 covers the entire outside of the adhesive sound insulating portion 14 other than the skin adhesive portion 14a, and forms the exterior of the voice detecting portion 10. That is, the outer cover member 15 is a container-like member whose one surface is in an open state, and an inner cover member 13 that encloses the NAM propagation unit 12 and the microphone 11 is housed inside thereof, and adhesive sound insulation. The portion 14 (urethane elastomer) is filled with almost no gap.

Here, the adhesive sound insulation unit 14 causes the skin contact surface 12a of the NAM propagation unit 12 to be in close contact with the skin surface 1a when the skin adhesive unit 14a adheres to the skin surface 1a, and the voice detection unit 10 is It is held in a state where it is adhered to the human body (mounted state). In addition, by comprising the NAM propagation part 12 with urethane elastomer (adhesive soft member), both the skin contact surface 12a and the skin adhesion part 14a of the NAM propagation part 12 adhere to the skin surface 1a, The voice detection unit 10 is preferable because it is more firmly held against the human body.
The adhesive sound insulation part 14 covers the entire outside of the inner cover member 13 and propagates in the air by the skin adhesion part 14a adhering to the skin surface 1a around the skin contact surface 12a of the NAM propagation part 12. It also functions as a sound insulating material that prevents disturbance sound from entering the microphone 11.

Next, the configuration of the signal processing unit 20 will be described with reference to FIG.
The signal processing unit 20 includes an amplifier 21, a high-pass filter 22, an A / D conversion unit 23, a wireless communication unit 24, an antenna 25, and a battery 26.
The amplifier 21 amplifies the inaudible murmur voice signal (the voice signal obtained by the microphone 11) transmitted from the voice detection unit 10.
The high-pass filter 22 performs a high-pass filter process on the inaudible murmur voice signal (the voice signal obtained by the microphone 11) transmitted from the voice detection unit 10, and outputs the processed signal to a subsequent circuit. Is.
Here, the amplifier 21 is disposed between the microphone 11 and the high-pass filter 22 in the sound detection unit 10 and amplifies a signal of a non-audible tweet sound before being input to the high-pass filter 22.
This makes it possible to amplify a signal of a non-audible murmur voice having a low volume level to a level sufficient as an input signal of the high-pass filter 22.
The high-pass filter 22 shown in FIG. 1C is a very simple CR circuit (differential circuit) composed of a capacitor and a resistance element. For example, it is conceivable that the capacitor capacity is about 0.1 μF and the resistance value is about 1.6 kΩ (cutoff frequency≈1 kHz).

The A / D converter 23 converts the inaudible audio signal (analog signal) amplified by the amplifier 21 and filtered by the high-pass filter 22 into a digital signal at a predetermined sampling frequency. For example, the A / D converter 23 performs A / D conversion at a sampling frequency of about 8 kHz.
The wireless communication unit 24 wirelessly transmits the inaudible audio signal digitized by the A / D conversion unit 23 to an external device having a communication function through the antenna 25. For example, a digital audio signal is transmitted to an external device in accordance with a well-known Bluetooth communication standard.
The battery 26 supplies power to each device (the amplifier 21, the A / D conversion unit 23, and the wireless communication unit 24) constituting the signal processing unit 20. The battery 26 also supplies power to the microphone 11 when the microphone 11 of the sound detection unit 10 requires driving power (for example, a bias condenser microphone). Of course, when the microphone 11 does not require driving power (for example, a back electret condenser microphone or the like), power supply to the sound detection unit 10 side is not required.

FIG. 2 is a schematic diagram showing a state in which the NAM microphone X is worn on the human body.
As shown in FIG. 2, the voice detection unit 10 is configured so that the skin contact surface 12 a of the NAM propagation unit 12 is in close contact with the skin surface on the thoracic papillary muscle immediately below the mastoid process of the skull in the lower part of the auricle. Then, it is attached to the human body (adhered by the adhesive sound insulation unit 14). Thereby, inaudible murmur sound (NAM) generated in the vocal tract efficiently propagates from the flesh of the body to the NAM propagation unit 12 without the bones or the like becoming an obstacle.
Further, the signal processing unit 20 is attached to the human body when the ear mounting housing 27 that is the housing engages the ear.
Thereby, in the hands-free state, the signal of the inaudible murmur voice detected by the voice detection unit 10 and subjected to the high puff filter processing is wirelessly transmitted from the signal processing unit 20 to the external device. Therefore, if the external device that can communicate with the signal processing unit 20 includes, for example, a voice recognition function based on a signal of a non-audible murmur voice and an automatic control function that controls the operation of the own device according to the recognized voice. The wearer of the NAM microphone X can remotely control the external device in a hands-free state and by uttering a non-audible murmur voice that does not leak sound around.
Further, if a device that receives a signal of non-audible murmur voice from an external device and outputs it as a voice is added to the signal processing unit 20, a caller using a non-audible murmur voice is obtained. In this case, for example, the wireless communication unit 24 is provided with a function of receiving a signal of a non-audible murmur voice from an external device. Further, the signal processing unit 20 includes a D / A conversion unit that converts an audio signal (digital signal) received by the wireless communication unit 24 into an analog signal, and an amplifier that amplifies the audio signal (analog signal) after the D / A conversion. And a speaker (earphone) for outputting the amplified audio signal. As a result, the NAM microphone X becomes a caller using a non-audible tweet voice.

By the way, the non-audible murmuring sound (NAM) collected as the meat conduction sound and the audible whispering sound collected as the air conduction sound are both unvoiced sounds (sounds without accompanying vocal cord vibration). However, as described above, it is difficult to recognize the utterance content of non-audible tweet speech simply by amplification (it is difficult to hear), and audible tweet speech is relatively easy to recognize the utterance content.
FIG. 3 shows a voiceprint (a) of a non-audible tweet voice (NAM) collected by a conventional NAM microphone (without filtering) and a voiceprint (b) of an audible whisper voice collected as an air conduction sound by a general microphone. ).
Comparing FIG. 3 (a) and FIG. 3 (b), the non-audible murmur voice (a) has a signal intensity of a frequency component of 750 Hz or less compared to the whisper voice (b) in which the utterance content is relatively easy to recognize. There is a strong tendency. This is a signal component (S) in which the signal component exceeding 750 Hz mainly contributes to the identification of the utterance content in the signal of the non-audible whispering sound, and the signal component of 750 Hz or less mainly does not contribute to the identification of the utterance content. It is expected to be component (N).

From the above, it is expected that the signal-to-noise ratio will be improved if the high-pass filter 22 (that is, the low-cut filter) removes a frequency component of 750 Hz or less from the inaudible murmur voice signal collected as the meat conduction sound. .
In fact, it has been found that when a non-audible murmur voice is collected by the NAM microphone X, a voice signal that is easy to hear (easy to recognize) can be obtained by simply amplifying it. The ease of listening is comparable to the audible whispering sound.
In addition, the sound obtained by the NAM microphone according to the present invention includes a sound with an unnatural intonation and a false sound that is not originally uttered, such as a normal sound (voiced sound) converted based on a sound conversion model. There is no stability.
Furthermore, even when the high-pass filter 22 is incorporated in a small device such as a mobile phone or a body-worn device, the weight and volume of the device are hardly increased, and the continuous use time is shortened. There is nothing.
In addition, the NAM microphone X does not require processing with a high calculation load unlike the signal conversion processing based on the voice conversion model, so that there is no delay in audio signal transmission. As a result, it is possible to realize a smooth conversation by non-audible murmur voice and remote control of the device.

Next, a first evaluation experiment and a second evaluation experiment, which are experiments evaluating the relationship between the characteristics of the high-pass filter employed in the NAM microphone X and the microphone performance, will be described.
In each of the first and second evaluation experiments, 22 subjects (10 males and 12 females) each have 22 types of sample sounds (first to first) with different contents of sound source and signal processing (high-pass filter processing). 22 sample voices) and according to a sense (evaluation result) obtained by the listening, answers are made for predetermined evaluation items.
Here, the first to tenth sample sounds are cut out of meat conduction sounds when a predetermined utterer utters a predetermined sample text (including sentences and words) with a non-audible muttering sound (NAM). This is a non-audible whispered sound after filtering recorded by a NAM microphone X having a high-pass filter with an off frequency of 200, 400, 600,..., 2000 Hz (in 200 Hz increments) and a slope characteristic of −120 dB / oct.
In addition, the 11th to 20th sample sounds are meat conduction sounds when a predetermined utterer utters the sample text with a non-audible muttering sound, with cut-off frequencies of 200, 400, 600,. And slope characteristics are -23.0, -18.6, -16.8, -15.6, -15.0, -14.4, -13.9, -13.6, respectively. , 13.2 and -13.0 dB / oct, a non-audible whispered sound after filtering recorded by the NAM microphone X provided with a high-pass filter.
The 21st sample sound is the audible whisper sound recorded by a normal microphone (microphone for collecting air conduction sound).
The twenty-second sample sound is a non-audible whisper sound (NAM) recorded by a conventional NAM microphone that does not perform high-pass filter processing.
The sample text is a newspaper article having about 20 to 30 words.
Each sample sound was recorded as digital sound (PCM sound) having a quantization bit number of 16 bits and a sampling rate of 8 kHz, and played back.

<First evaluation experiment (naturalness evaluation experiment)>
In the first evaluation experiment, each subject listened to 11 pairs of sample sounds of 2 types arbitrarily selected from 22 types of sample sounds. Was selected as a natural conversational voice. In this first evaluation experiment, it can be said that the sample voice selected as natural by more subjects has high naturalness.
FIG. 4 is a graph showing the results of the first evaluation experiment (naturalness evaluation experiment). For each of the 22 types of sample signals, the subject sample signal is compared with the other sample signals to be compared. Represents the ratio selected as natural (an alternative selection ratio).
As can be seen from FIG. 4, the non-audible tweet voice (the 21st sample voice v21) that is not subjected to the high-pass filter processing has a rate (selectivity) at which it is evaluated as natural, which is less than 50%. The rate at which the fifteenth sample voice v15 (high-pass filter processing with a cutoff frequency of 1000 Hz and a slope characteristic of -15.0 dB / oct) is evaluated as natural is over 80%. Here, since the selection rate of the audible whisper sound (22nd sample sound v22) is over 90%, the naturalness is comparable to the audible whisper sound simply by applying the high-pass filter process to the non-audible murmur sound. It can be seen that the naturalness is improved to a certain extent.
Further, as can be seen from FIG. 4, non-audible muttering voices (fourth to sixth, fourteenth to sixteenth sample voice signals v4) subjected to high-pass filter processing with a cutoff frequency of 800 to 1200 Hz, regardless of the slope characteristics. ˜v6, v14 to v16) exceeds 60%, and an improvement in naturalness is observed.
Further, as can be seen from FIG. 4, a signal having a relatively gentle slope characteristic (particularly, the 14th to 20th sample signals) is higher than a signal having a steep (−120 dB) slope characteristic in high-pass filter processing. Suitable for improving naturalness.

<Second evaluation experiment (word recognition accuracy evaluation experiment)>
In the second evaluation experiment, the subject listens to an arbitrary selection of 22 types of sample voices that are read-out voices of newspaper articles (that is, meaningful sentences), and the word recognition accuracy (recognized by the subject). The correct answer rate of words was evaluated.
FIG. 5 is a graph showing the results of the second evaluation experiment (word recognition accuracy evaluation experiment), and for each of the 22 types of sample signals (reading speech of newspaper articles), the subject's word recognition accuracy (correction of recognized words). Rate).
As can be seen from FIG. 5, the non-audible murmur voice (the twenty-first sample voice v21) not subjected to the high-pass filter processing has a word recognition accuracy of about 85%, whereas the fourteenth sample voice v14 (cutoff frequency). The high-pass filter processing (800 Hz, slope characteristic −15.6 dB / oct) has a word recognition accuracy of about 90%.
Further, as can be seen from FIG. 5, the non-audible murmur sound subjected to the high-pass filter processing having a relatively gentle slope characteristic (-23.0 to −13.9 dB / oct) in the range of the cutoff frequency of 200 to 1400 Hz. (Fourteenth to seventeenth sample voice signals v14 to v17) have a word recognition accuracy equal to or higher than that of a non-audible murmur voice that is not subjected to high-pass filter processing.
Similarly, when the slope characteristic is steep (−120 dB / oct), the non-audible murmur sound (first to third sample audio signals v1 to v3) subjected to the high-pass filter processing with a cutoff frequency of 200 to 600 Hz is obtained. Compared with a non-audible tweet voice that is not subjected to high-pass filter processing, a word recognition accuracy equal to or higher than that can be obtained.
As described above, in the NAM microphone X, the cutoff frequency of the high-pass filter 22 is set to about 800 to 1400 Hz (particularly about 800 to 1000 Hz), and the slope characteristic is set to about −14 to −16 dB / oct. It turns out that it is possible to collect sounds that are easy to hear (easy to recognize words).

In the embodiment described above, the NAM microphone X (wireless type microphone) that wirelessly transmits (transmits) the inaudible murmur sound after the high-pass filter processing has been described. However, the present invention is not limited to this.
For example, the A / D converter 23 and the wireless communication unit 24 are removed from the NAM microphone X shown in FIG. 1, and an output terminal connected to the audio signal (analog) input terminal of the external device is provided instead. A wired NAM microphone is also conceivable.
The high-pass filter 22 may be configured by a known circuit other than the CR circuit as long as it has a cutoff frequency of about 750 Hz to 1400 Hz.
Further, the slope characteristic of the high-pass filter 22 is not limited to about −16 dB / oct to −14 dB / oct.

  The present invention can be used for a microphone for collecting non-audible murmur voices.

1 is a schematic configuration diagram (partial block diagram) of a NAM microphone X according to an embodiment of the present invention. The schematic diagram showing the state with which NAM microphone X was mounted | worn with the human body. The figure showing the voiceprint of a non-audible murmur voice and the voiceprint of an audible whisper voice. The graph showing the result of the 1st evaluation experiment (natural nature evaluation experiment) of NAM microphone X. The graph showing the result of the 2nd evaluation experiment (word recognition accuracy evaluation experiment) of NAM microphone X.

Explanation of symbols

X: Microphone 10 for collecting an inaudible murmur sound according to an embodiment of the present invention 10 ... Audio detecting unit 11 ... Microphone 12 ... NAM propagation unit 13 ... Inner cover member 14 ... Adhesive sound insulating unit 15 ... Outer cover member 20 ... Signal processing unit 21 ... Amplifier 22 ... High-pass filter 23 ... A / D conversion unit 24 ... Wireless communication unit 25 ... Antenna 26 ... Battery 27 ... Ear mounting case 30 ... Signal line

Claims (4)

A non-audible muttering voice collecting microphone for collecting a non-audible muttering voice that is a voice generated by a breathing sound that is not accompanied by a vocal cord vibration generated in a human vocal tract,
A soft member that propagates the inaudible murmur sound by being in close contact with the skin surface of the human body;
A microphone that converts the inaudible murmur sound propagating through the soft member into an electrical signal;
A high-pass filter that performs high-pass filter processing on the inaudible tweet signal obtained by the microphone;
A microphone for collecting a non-audible tweet sound characterized by comprising:   The inaudible muttering voice collecting microphone according to claim 1, wherein a cutoff frequency of the high-pass filter is approximately 800 Hz to approximately 1400 Hz.   The inaudible muttering voice collecting microphone according to claim 2, wherein a slope characteristic of the high-pass filter is approximately -16 dB / oct to approximately -14 dB / oct.   The inaudible tweet voice collecting microphone according to any one of claims 1 to 3, further comprising an amplifier that amplifies the signal of the inaudible tweet voice between the microphone and the high-pass filter.

Images