JP5391374B2 - Microphone device - Google Patents

Description

  The present invention relates to a microphone device using sound conducted through a human body.

  In recent years, the acquisition of voice (bone conduction sound) conducted through the human body, particularly the skeleton, has been attracting attention as a technique that is less susceptible to ambient noise and enables voice calls and voice input under noise. However, in obtaining bone conduction sound, (1) the clarity is low because there are few high frequency components important for sound expression, and (2) the frequency characteristic of bone conduction speech is high in the low range and low in the high range. Sound quality is poor, (3) clarity varies greatly depending on individual differences in the external auditory canal and how the pickup unit is attached, (4) it is easy to pick up abnormal sounds such as friction of cords (signal lines), (5) Easy to pick up motility noise, (6) Poor efficiency and S / N ratio, (7) Sound quality inherent to bone-conducted speech due to very complicated voice transmission from each part such as vocal cords, vocal tract or mouth It is said that there is a disadvantage that occurs.

The above (1) to (6) are problems caused by the inadequate conventional method for obtaining bone conduction sound. In order to solve the problems (1) to (6), the contact of the pickup element is provided with an adjustment function for relaxing the difference in the distance from the detection part of the human body, and correction for individual differences in the human body is easily performed. A technique for improving the fit or wearability is disclosed (for example, see Patent Document 1). Furthermore, according to the technique described in Patent Document 1, since the contact is in contact with the human body, the clarity and S / N ratio of the bone conduction speech recognition are improved, and the bone conduction speech is recognized well. can do.
JP 2002-262377 A

  By the way, the above (7) is an essential problem of bone conduction sound, and in order to obtain highly reproducible audio information equivalent to air conduction sound from bone conduction sound, appropriate correction is necessary. is there. Patent Document 1 does not describe such correction processing.

  The present invention has been made in view of such a situation, and makes it possible to obtain sound information having high reproducibility equivalent to air conduction sound from acquired bone conduction sound.

A microphone device according to one aspect of the present invention has a plate-like detection unit that is a piezoelectric element that detects vibration of a living body surface that is in close contact with the plate-like one surface and converts the vibration into an electric signal, and the detection unit Based on the electrical signal corresponding to the vibration of the living body surface by the impulse-like sound uttered from the living organ that is output from the living body , the living body from the living body's vocal organ to the position where the detecting unit of the living body surface is attached Applying a transfer function, a calculation unit that calculates the inverse transfer function, and a filter that performs calculation using the inverse transfer function calculated by the calculation unit to the electrical signal output from the detection unit A signal processing unit to perform, and an output unit that outputs the electrical signal filtered by the signal processing unit to the outside. In addition, it collects air-conducted sound uttered from a living body's vocal organs and propagates through the space, and converts it into an electrical signal; an electrical signal to which the filter is applied in the signal processing unit; and an output from the sound collection unit A comparison operation unit that compares the calculated electric signal and calculates a correction amount of the inverse transfer function calculated by the operation unit based on the comparison result. The calculation unit corrects the inverse transfer function based on the correction amount calculated by the comparison calculation unit.

  In one aspect of the present invention, bone conduction vibration information is corrected using a reverse transfer function from the user's vocal organ to the position where the body surface detection unit is attached. High sound quality can be obtained.

  As described above, according to one aspect of the present invention, voice information having high reproducibility equivalent to air conduction sound can be obtained from the acquired bone conduction sound.

  Hereinafter, examples of embodiments of the present invention will be described with reference to the accompanying drawings.

  The embodiment described below is a specific example of a preferred embodiment for carrying out the present invention, and therefore various technically preferable limitations are given. However, the present invention is not limited to these embodiments unless otherwise specified in the following description of the embodiments. Therefore, for example, the materials used in the following description, the amounts used, the processing time, the processing order, and the numerical conditions of each parameter are only suitable examples, and the dimensions, shapes, and arrangements in the drawings used for the description The relationship and the like are also schematic showing an example of the embodiment.

  The present invention uses a voice conducted through a human body as a technique for recognizing a voice produced by a living body such as a human, and particularly relates to a microphone device that acquires a voice (bone conduction sound) conducted through a skeleton. The microphone device is roughly divided into a sensor capable of measuring fine vibrations generated on the body surface by sound generation up to a high frequency range, and a device for correcting an audio signal output from the sensor.

  In one embodiment of the present invention, an element having a piezoelectric effect that generates an electric charge according to strain, that is, a polyvinylidene difluoride (PVDF) film, which is a piezoelectric element, is used for the sensor. The polyvinylidene fluoride (hereinafter referred to as “PVDF”) film is used to form a PVDF bone conduction sound sensor, and the PVDF bone conduction sound sensor is fixed at an arbitrary position on the body surface. Bone conduction sound is acquired by detecting vibration of the body surface.

FIG. 1 is a diagram showing a usage example of a PVDF bone conduction sound sensor according to an embodiment of the present invention.
The illustrated example is an example in which the PVDF film of the PVDF bone conduction sound sensor 1 is brought into close contact with the user's forehead and fixed with fixing means 2 such as a rubber band. A voice emitted from the user is detected by the PVDF bone conduction sound sensor 1 fixed to the forehead.

  FIG. 2 is a perspective view showing a schematic configuration of the PVDF bone conduction sound sensor shown in FIG. Moreover, FIG. 3 is sectional drawing which follows the XX line of the PVDF bone-conduction sound sensor 1 shown in FIG. In FIG. 3, the description of the fixing means 2 is omitted.

  As shown in FIGS. 2 and 3, the PVDF bone conduction sound sensor 1 is configured to include a PVDF film 11, a buffer material 12, an insulating plate 13, and fixing means 2. Further, an electrode 14 is provided near one end of the PVDF film 11, and a lead wire 15 is connected to the electrode 14. The lead wire 15 is connected to an amplification unit 21 described later.

  The PVDF film 11 is formed in a substantially rectangular or square plate shape so that one surface thereof is in close contact with the body surface (for example, the surface of the forehead). When the shape of the PVDF film 11 is a rectangular plate shape, it is known that the longitudinal direction is the detection direction of bone conduction vibration. In the example shown in FIG. 2, the detection direction of the biological vibration of the PVDF film 11 is the longitudinal direction of the PVDF film 11 along the body surface, and thus is the short direction of the band-shaped fixing means 2, that is, the longitudinal direction V1 of the forehead. .

  The cushioning material 12 is used to uniformly press the PVDF film 11 against the body surface and acquire high-quality sound, and is sandwiched between the insulating plate 13 and the PVDF film 11. The shape of the cushioning material 12 is approximately plate-like, and a sponge (foam) having a sponge-like structure is used for the cushioning material 11 so that the pressing force from the fixing means 2 is not directly applied via the insulating plate 13. It is done. The area (size) of the plane portion is preferably equal to or larger than the area (size) of the plane portion of the PVDF film 11 in order to uniformly press the PVDF film 11 against the body surface. As the material of the buffer material 12, for example, a nitrile rubber (NBR), an ethylene vinyl acetate copolymer resin (EVA) used for a cosmetic puff, or an elastic material mainly composed of an organic polymer such as plain rubber is used. .

  In addition, you may make it press the PVDF film 11 against a body surface by fixing means, such as a rubber band, without providing the shock absorbing material 12. FIG. However, it has been confirmed by experiments by the present inventors that the sound reproducibility is better when the buffer material 12 is provided than when the buffer material 12 is not used. In addition, the cushioning material is not limited to a sponge but may be a material having another composition or structure. However, the sponge has a higher frequency reproducibility than other materials. Furthermore, the frequency characteristics of the sound detected by the PVDF film 11 vary depending on the thickness of the sponge.

  The insulating plate 13 is provided between the side on which the PVDF film 11 and the sponge 12 are provided and the fixing means 2 side, and prevents conduction of electricity and heat. In the present embodiment, a rubber plate is used as the insulating plate 13, but is not limited to this example.

  The fixing means 2 applies force in the vertical direction of the flat portion of the PVDF film 11 in order to uniformly press the PVDF film 11 against the body surface. It is only necessary that the PVDF film 11 can be in close contact with the epidermis by the fixing means 2, and a strong pressing force is unnecessary. In this example, a rubber band is used, but the present invention is not limited to this example. For example, the PVDF film 11 may be brought into close contact with the body surface using a hat, a helmet, or goggles. Furthermore, you may make it apply | coat an adhesive agent to the surface which closely_contact | adheres to the forehead of a PVDF film, or affixes an adhesive tape, and makes a PVDF film and an amount | piece adhere. However, the experiment by the applicant of the present application has confirmed that the sensitivity of detecting fine vibrations on the body surface is better when the adhesive and the adhesive tape are not used than when they are used. .

  Most conventional sensors that detect bone-conduction vibrations output the deformation of the internal structure of the sensor caused by vibration caused by biological vibrations instead of electric charges (charge signals). On the other hand, the PVDF bone conduction sound sensor 1 according to the present embodiment, for example, closes the PVDF film 11 to the skin of the forehead, and outputs the distortion generated in the skin due to the bone conduction sound instead of the direct charge (charge signal). . Thereby, it is possible to have flat output characteristics with no undulations in a wide frequency band including a high frequency regardless of the resonance characteristics of the PVDF bone conduction sound sensor 1 itself.

  Further, since the detection direction of the biological vibration is a direction along the body surface, it is difficult to be affected by sound pressure due to external noise (applied in a direction perpendicular to the body surface).

  Furthermore, since the impedance between air and the body surface is greatly different, the skin is hardly distorted by noise from outside the living body. Therefore, external noise can be blocked by the PVDF bone conduction sound sensor 1 alone without requiring a special filter circuit.

FIG. 4 is a diagram showing another example of use of the PVDF bone conduction sound sensor according to one embodiment of the present invention.
When the bone conduction vibration of the voice emitted from the user's voice organ is detected with the forehead, it has been confirmed by the inventors' experiment that the vibration direction is the lateral direction along the skin of the forehead. Therefore, as shown in FIG. 4, the PVDF bone sound sensor 1 is fixed so that the longitudinal direction of the PVDF bone sound sensor 1 (PVDF film) is along the longitudinal direction of the belt-like fixing means 2. Thus, by making the detection direction of the bone conduction vibration of the PVDF bone conduction sensor 1 to be the longitudinal direction of the fixing means 2, that is, the lateral direction V2 of the forehead, the PVDF bone conduction sensor 1 shown in FIG. Fine vibrations can be detected.

Here, with reference to FIGS. 5-8, a PVDF bone-conduction sound sensor is evaluated based on the result obtained by the experiment in a noisy environment.
The PVDF bone conduction sensor used in the experiment is “DT1-028K / L (30 × 12 mm, thickness 28 μm)” manufactured by Tokyo Sensor Co., Ltd. on the PVDF film 11, sponge as the buffer material 12, and rubber plate as the insulating plate 13. The sponge (buffer material 12) was sandwiched between the PVDF film 11 and the rubber plate (insulating plate 13). The PVDF bone conduction sound sensor was fixed using a rubber band as the fixing means 2. Further, a plastic plate was attached to the rubber plate (insulating plate 13), and the plastic plate was disposed between the rubber plate (insulating plate 13) and the rubber band (fixing means 2).

  FIG. 5 is a diagram illustrating an example of a result of a short-time fast Fourier transform (short-time FFT (FFT)) of a signal obtained from the PVDF bone conduction sound sensor under white noise. The horizontal axis represents time [s], and the vertical axis represents frequency [Hz] and power spectrum. The display of the power spectrum of the original data of the application is not displayed in color (chromatic color) in the figure, but in reality, it is displayed in stages such as red, green, and blue from the strongest to the weakest. Yes. The measurement was performed by a method in which a signal output from the PVDF bone conduction sound sensor is detected in a state where the subject does not speak anything under white noise of 90 dB, and the detected signal is subjected to FFT processing for a short time. White noise is noise including all frequencies (over a wide band).

  In general, in the case of a microphone that detects air conduction sound, it turns red in the entire frequency band. However, in the example shown in FIG. 5, the detection of external noise (sound conducted in the air) is suppressed, and it turns green in a wide area. It has become. That is, in the case of the PVDF bone conduction sound sensor, noise other than the band near 1000 Hz to 3000 Hz is suppressed by about 30 dB, and it can be seen that the PVDF bone conduction sound sensor itself functions as a bandpass filter. In general, the frequency band necessary for obtaining voice is about 200 Hz to 5000 Hz.

  FIG. 6 is a diagram showing an example of a short-time FFT result of an audio signal measured under white noise, where (a) is a microphone and (b) is a PVDF bone conduction sensor. The horizontal axis represents time [s], the vertical axis represents frequency [Hz] and the power spectrum, and the original data is displayed in color as in FIG. The measurement was performed by a method of detecting a signal output from the PVDF bone conduction sensor when the subject reads a news manuscript under 90 dB white noise and performing FFT processing on the detected signal for a short time. The subject's voice is subjected to a band pass filter of 200 Hz to 5000 Hz.

  As shown in FIG. 6A, in the case of a microphone, external noise is picked up even in a frequency band of about 5000 Hz or more under white noise and is displayed in red. On the other hand, in the case of the PVDF bone conduction sound sensor, as shown in FIG. 6 (b), although the speech intelligibility is low, noise in a band exceeding about 3000 Hz is suppressed, and the conversation contents of the subject can be heard in a fragmentary manner. It has become.

  FIG. 7 is a diagram illustrating an example of a short-time FFT result of an audio signal measured under construction site noise, where (a) is a microphone and (b) is a PVDF bone conduction sensor. The horizontal axis represents time [s], the vertical axis represents frequency [Hz] and the power spectrum, and the original data is displayed in color as in FIG. The measurement was performed by a method in which a signal output from the PVDF bone conduction sensor when a subject reads a news manuscript under a construction site noise of 85 dB is detected, and the detected signal is subjected to FFT processing for a short time. As described above, the subject's voice is subjected to a band-pass filter of 200 Hz to 5000 Hz.

  As shown in FIG. 7A, in the case of a microphone, noise of approximately 1000 Hz or less is picked up over the entire time. On the other hand, FIG. 7B shows that in the case of the PVDF bone conduction sensor, noise of approximately 1000 Hz or less is cut off over the entire time.

  FIG. 8 is a diagram showing an example of a waveform of an audio signal measured under the construction site noise of FIG. 7, where (a) is a microphone and (b) is a PVDF bone conduction sensor. As can be seen from FIGS. 8A and 8B, in the case of the PVDF bone-conducted sound sensor, the waveform of the voice changes with the conversation content, and the conversation content of the subject can be clearly heard.

  As described above, the PVDF film 11 of the PVDF bone conduction sensor 1 only needs to be in close contact with an arbitrary position on the body surface of the user, and a strong pressing force is not required. Therefore, the PVDF film itself has a gentle touch. Can be processed into a soft and thin shape. For this reason, it is easy to wear on the user's body surface and can be worn for a long time.

  In addition, since the detection direction of biological vibration (bone conduction vibration) is a direction along the body surface, it is not easily affected by sound pressure due to external noise (applied in a direction perpendicular to the body surface) and is strong in a noise environment.

  Further, since one surface of the PVDF film 11 of the PVDF bone conduction sound sensor 1 is brought into close contact with the epidermis, there is no influence of resonance derived from the sensor structure, and a flat frequency characteristic up to a high range is obtained.

  The measurement data shown in FIGS. 5 to 8 is an example in the case where the shock absorbing material 12 composed of a sponge (foam) having a spongy structure is used for the PVDF bone conduction sound sensor. The frequency range and output level that can be acquired vary depending on the type. However, even if the material changes, the difference is not so great if it is sponge-like.

On the other hand, when a hard plastic plate was inserted instead of a sponge (foam) having a spongy structure, the results of the present inventors showed that the frequency range was considerably narrowed. However, when a plastic plate is used, the external noise also decreases at the frequency of the voice signal (200 to 4000 Hz), so the content of the conversation is important and the voice quality is not a problem (not used for voice recognition, etc.) ) Is useful. On the other hand, when a sponge is used, it can be said that it is suitable for use in speech recognition and the like because the measurable frequency range is wide as long as external noise can be eliminated.
That is, when a sponge is used, the frequency range is about 8000 Hz or less, which is suitable for voice recognition. On the other hand, when a plastic plate is used, the frequency range is about 4000 Hz or less, and external noise is small in the conversation area, which is suitable for speech recognition.

  By the way, the bone conduction sound and the air conduction sound have different transmission paths, and therefore have superior (easy to conduct) frequency components. Therefore, even if the bone conduction vibration is detected with high quality by the PVDF bone conduction sound sensor and the detected signal is reproduced, the sound quality is different from the air conduction sound. Usually, animals including humans communicate with each other by air conduction sound. Therefore, if the bone conduction sound is accurately reproduced, it cannot function as a high-quality microphone (also referred to as “Hi-Fi microphone”) that faithfully reproduces the original sound uttered by the human vocal organs. It is necessary to convert the bone conduction vibration into an electric signal having a frequency component equivalent to the air conduction sound.

  Therefore, in the present invention, for example, by using a pseudo impulse sound generated by tongue hitting or the like, a transfer function from a vocal organ of the human body to a place where the PVDF bone conduction sensor on the body surface such as the forehead is fixed is obtained, and by the inverse transfer function, The electrical signal corresponding to the bone conduction vibration acquired with high quality is corrected and output. In addition, when the surrounding environment is relatively quiet, a learning circuit that always corrects the transfer function and its inverse transfer function from voice information and bone conduction vibration information collected by a normal microphone is also prepared.

  Next, a microphone device to which the PVDF bone conduction sound sensor according to one embodiment of the present invention is applied will be described.

FIG. 9 is a block diagram showing a schematic configuration of a microphone device to which the above-described PVDF bone conduction sound sensor is applied.
A microphone device 20 shown in FIG. 9 includes a PVDF bone conduction sound sensor 1, an amplification unit 21, a signal processing unit 22, an output unit 23, a switch 24, a microphone 25, an amplification unit 26, a switch 27, A transfer function calculation processing unit 28 is included. Further, the program memory 30, the volatile memory 31, the nonvolatile memory 32, the control calculation unit 33, the comparison calculation unit 34, and the operation unit 35 are included. The blocks are connected to each other via a bus 29 so that data can be transmitted / received to / from each other.

  As described above, the PVDF bone conduction sound sensor 1 is composed of the PVDF film 11 (an example of the detection unit described in the claims) and other parts, and is attached to the body surface of the user, for example, the forehead. The bone conduction vibration of the body surface by the sound emitted from the voice organ of the user is detected by the PVDF film 11, the electric charge corresponding to the detected vibration is generated and output to the amplifying unit 21.

  The amplifying unit 21 converts the charge information input from the PVDF bone conduction sound sensor 1 into an electric signal (voltage signal) and outputs the electric signal (voltage signal) to the signal processing unit 22 or the switch 24. As the amplifying unit 21, a so-called charge amplifier that converts a charge signal into a voltage signal can be used.

The signal processing unit 22 includes a filter circuit, applies a filter to the electric signal u (t) input from the amplification unit 21, and outputs the filtered output audio signal y (t) to the output unit 23. It is. The filter circuit performs filtering based on the inverse transfer function G (s) ⁻¹ with respect to the transfer function G (s) from the user's voice organ to the position where the PVDF bone conduction sound sensor 1 on the human body surface is attached. The inverse transfer function G (s) ⁻¹ used for the calculation is read from the nonvolatile memory 32 and used. Note that t represents time, s represents a complex number, t region represents a time region, and s region represents a complex region. As the signal processing unit 22, a digital signal processor (DSP) can be used in addition to a general-purpose processor.

Here, the filtering process performed by the signal processing unit 22 will be described. First, a Laplace transform is performed on the input signal u (t) in the t region to calculate a function U (s) in the s region, and the function U (s) is convolved with the inverse transfer function G (s) ⁻¹ to integrate the function. Y (s) is obtained. The function Y (s) is subjected to inverse Laplace transform to obtain an output audio signal y (t) represented by a time domain function. That is, the signal processing unit 22 performs two major processes.
(1) Y (s) = G (s) ⁻¹ · U (s)
(2) Inverse Laplace transform Y (s) → y (t)

  The output unit 23 is an interface that outputs an output audio signal supplied from the signal processing unit 22 to an external speaker, a recorder, or a voice recognition device.

  The switch 24 is switching means for switching between an open state and a closed state by a control signal from the control calculation unit 32. When the switch 24 is in the closed state, the electrical signal output from the amplifying unit 21 is input to the transfer function calculation processing unit 28 via the switch 24.

The microphone 25, the amplifying unit 26, and the switch 27 include a transfer function G (s) from the user's voice organ to the position where the PVDF bone conduction sensor 1 on the surface of the human body is attached, and an inverse transfer function G (s) ^{− 1} is a circuit provided for acquiring a signal for correcting ¹ ;

  The microphone 25 is an example of a sound collecting unit described in the claims, and air conduction vibration (sound pressure) of a sound emitted from a user's utterance organ to which the PVDF bone conduction sound sensor 1 is attached to the body surface. , And an electric signal (voltage signal) corresponding to the detected air conduction vibration is generated and output to the amplifying unit 26. The microphone 25 may be attached to an arbitrary place on the user's body, or installed near the user's mouth only in the correction mode described later as a completely separate configuration from the microphone device 20. You may make it connect with the amplifier 26 of the microphone apparatus 20. FIG.

  The amplifying unit 26 amplifies the electric signal input from the microphone 25 and outputs the amplified electric signal to the switch 27. The amplification unit 26 may not be provided when the signal level of the electric signal input from the microphone 25 is sufficiently high.

  The switch 27 is switching means that switches between an open state and a closed state in accordance with a control signal from the control calculation unit 33. When the switch 26 is in the closed state, the electrical signal m (t) output from the microphone 25 is input to the transfer function calculation processing unit 28 via the switch 27.

  The transfer function calculation processing unit 28 is an example of a calculation unit described in the claims, and uses the Laplace transform that converts the input function of the t region into a function of the s region. A transfer function representing the dynamic characteristics of the process used and the process using the microphone 25 is obtained. As the transfer function arithmetic processing unit 28, a digital signal processor (DSP) can be used in addition to a general-purpose processor.

For example, in the present embodiment, U ₀ (s) is a function obtained by Laplace transform of a speech (original sound) signal u ₀ (t) uttered from a human vocal organ, and a PVDF bone on the body surface from the user's vocal organ. The transfer function up to the position where the sound sensor 1 is attached is G (s), and the function obtained by Laplace transforming the electric signal u (t) output from the PVDF bone sound sensor 1 and amplified by the amplifier 21 is U (s). ), The input / output relationship of the PVDF bone conduction sound sensor 1 is expressed by the following equation.
U (s) = G (s) · U ₀ (s) (1)

Here, when the signal u ₀ (t) of the voice uttered from the voicing organ is impulse-like, the signal u ₀ (t) becomes 1 when the Laplace transform is performed. Therefore, the following equation is derived from the equation (1). It is burned.
U (s) = G (s) (2)

  In other words, an impulse-like voice is uttered by the voicing organ, and at this time, the electrical signal u (t) input from the amplifying unit 21 is Laplace transformed, and a function U (s) in the s region is calculated. A transfer function G (s) to the position where the PVDF bone conduction sound sensor 1 on the body surface is attached is obtained. The obtained transfer function G (s) is temporarily stored in the volatile memory 31.

As an example of the impulse-like sound uttered by the voicing organ, a tongue-tapping sound can be cited. Compared with other sounds such as “A” and “O”, the tongue-sounding sound contains more frequency components, so that a more accurate frequency analysis of the voice uttered by the user can be performed. Of course, the transfer function from the user's vocal organs to a predetermined part of the body surface may be obtained from other vocal sounds other than the tongue hitting sound. Since it is easy to calculate the inverse transfer function G (s) ⁻¹ from the transfer function G (s), the transfer function calculation processing unit 28 calculates the transfer function G (s) and the inverse transfer function G. (S) ⁻¹ may also be calculated and temporarily stored in the volatile memory 31 or stored in the nonvolatile memory 32.

  The transfer function calculation processing unit 28 performs Laplace transform on the electrical signal m (t) output from the microphone 25 and amplified by the amplifier 26, and calculates a function M (s) in the s region. Then, the obtained function M (s) is temporarily stored in the volatile memory 31. Note that the sound collected by the microphone 25 and the sound corresponding to the bone motion vibration detected by the PVDF bone conduction sensor 1 are the same sound source of the same user for use in the comparison calculation in the comparison calculation unit 34 described later. Then, it is easy to perform the comparison process in the comparison calculation unit 34, which is convenient.

  The program memory 30 is a non-volatile storage unit such as a ROM (Read Only Member), and stores various programs and setting values that are read when the control calculation unit 33 performs control or calculation.

The volatile memory 31 is a volatile storage unit such as a RAM (Random Access Memory), and is used as a work area when the control calculation unit 33 performs control or calculation. The volatile memory 31 temporarily stores the transfer function G (s) and the inverse transfer function G (s) ⁻¹ calculated by the transfer function calculation processing unit 28.

The non-volatile memory 32 is a non-volatile recording means, and for example, a semiconductor memory such as a flash memory, a hard disk drive (HDD), a recording medium such as a DVD (Digital Versatile Disk), or the like is applied. The nonvolatile memory 32 stores the transfer function G (s) and the inverse transfer function G (s) ⁻¹ calculated by the transfer function calculation processing unit 28.

The control calculation unit 33 is configured by a processor such as a CPU (Central Processing Unit), for example, and reads out the program recorded in the program memory 30 to the volatile memory 31 and executes it, thereby controlling each block and processing data.・ Calculates. For example, although it has been described that the transfer function calculation processing unit 28 calculates the transfer function G (s) and the inverse transfer function G (s) ⁻¹ , the present invention is not limited to this example. A program for calculating the transfer function G (s) and its inverse transfer function G (s) ⁻¹ from an arbitrary signal is stored in the program memory 30, and the control calculation unit 33 executes the program, The transfer function G (s) of the input signal and its inverse transfer function G (s) ⁻¹ may be calculated.

The comparison calculation unit 34 is an example of the comparison calculation unit described in the claims, and is a main part of the learning function. In the comparison operation unit 34, the function Y (s) in the s region corresponding to the output audio signal y (t) of the signal processing unit 22 in the PVDF bone conduction sensor 1 system (the output function Y (s) when the input is an impulse. ) = G (s)) and a function M (s) in the s region corresponding to the output signal m (t) of the amplification unit 26 in the microphone 25 system. It is ideal that the function Y (s) and the function M (s) are the same, but actually, only the transfer function from the user's vocal organ to the position where the PVDF bone conduction sensor 1 on the body surface is attached. An error occurs. Therefore, when there is an error greater than or equal to a preset threshold between the function Y (s) and the function M (s), the correction amount ΔG (s) of the inverse transfer function G (s) ⁻¹ based on the error amount. ⁻¹ is calculated and supplied to the signal processing unit 22. In the signal processing unit 22 to which the correction amount ΔG (s) ⁻¹ is supplied, the reverse transfer function is corrected based on the correction amount ΔG (s) ⁻¹ input from the comparison calculation unit 34.

The operation unit 35 generates an operation signal according to a user's operation, and supplies the operation signal to the control arithmetic unit via an interface (not shown), and includes a push button and an operation key. When the user operates the operation unit 35, for example, a registration mode in which the reverse transfer function G (s) ⁻¹ from the user's vocal organ to the PVDF bone conduction sensor 1 is registered as an initial value, normal recording is performed. A recording mode and a correction mode for correcting bone conduction vibration information (reverse transfer function) are switched.

Next, the operation of the microphone device 20 configured as described above in the registration mode will be described.
When the initial value registration mode is selected, the control calculation unit 33 closes the switch 24. In the registration mode, when the PVDF bone conduction sound sensor 1 attached to the user's body surface (for example, forehead) detects bone conduction vibration of the body surface by the impulse-like sound emitted by the user, the PVDF bone conduction sound sensor 1 responds to the bone conduction vibration. Generate a charge signal. The charge signal is converted into an electric signal u (t) by the amplifier 21 and input to the transfer function arithmetic processing unit 28 via the switch 24. The transfer function calculation processing unit 28 transfers the transfer function G (s) and reverse transfer from the input electrical signal u (t) to the position where the PVDF bone conduction sound sensor 1 on the body surface is attached from the voice organ of the user. The function G (s) ⁻¹ is calculated. The inverse transfer function G (s) ⁻¹ is sent to the nonvolatile memory 32 and stored.

Next, the operation of the microphone device 20 in the recording mode will be described.
When the recording mode is selected, the control calculation unit 33 opens the switches 24 and 27. In parallel with this, the signal processing unit 22 stores the inverse transfer function G (s) ⁻¹ from the user's vocal organs registered in the registration mode to the PVDF bone conduction sensor 1 attached to the body surface in a nonvolatile manner. Read from the memory 32 and set to the internal filter.

When the PVDF bone conduction sound sensor 1 attached to the user's body surface (for example, the forehead) detects the bone conduction vibration of the body surface by the sound emitted by the user, a charge signal corresponding to the bone conduction vibration is generated. To do. This charge signal is converted into an electric signal u (t) by the amplifier 21 and input to the signal processor 22. The signal processing unit 22 calculates the output sound signal y (t) by applying a filter in which the inverse transfer function G (s) ⁻¹ is set to the input voltage signal u (t), and outputs the output sound signal y. (T) is output to the output unit 23. The output unit 23 outputs the output voice signal y (t) output from the signal processing unit 22 to a speaker, a recorder, a voice recognition device, or the like.

Thus, using the inverse transfer function G (s) ⁻¹ of the transfer function G (s) from the user's voice organ to the position where the PVDF bone conduction sensor 1 on the body surface is attached, the PVDF bone conduction sound is obtained. Since the bone conduction vibration information (electrical signal u (t)) detected by the sensor 1 is corrected, it is possible to obtain a sound that is close to the air conduction sound, that is, a sound that is faithful to the sound emitted from the vocal organ.

Further, since the transfer function G (s) and the inverse transfer function G (s) ⁻¹ can be derived from impulse-like sounds such as undertones as initial values, learning is performed in advance before using the microphone device. Easy to use.

  Note that the PVDF bone conduction sensor 1 and the amplifying unit 21 of the microphone device 20 described above are connected by wire or wirelessly, and each block including the amplifying unit 21 is housed in a housing, and the housing is, for example, the body of a viewer It may be attached at any position or stored in a bag or the like. Alternatively, at least the amplifying unit 21 may be integrated with the PVDF bone conduction sensor 1 in order to reduce noise. Furthermore, the functions of each block excluding the PVDF bone conduction sound sensor 1 or excluding the PVDF bone conduction sound sensor 1 and the amplification unit 21 may be provided in a portable device such as a cellular phone terminal.

  Next, the learning function of the microphone device 20, that is, the operation in the correction mode will be described with reference to FIG.

  First, in step S1, when the user operates the operation unit 35 to switch to the correction mode, the control calculation unit 33 detects that the correction mode has been switched and closes the switches 24 and 27. When this process ends, the process proceeds to step S2.

  In step S2, when the user utters an impulse-like sound such as a tapping sound, for example, the PVDF bone conduction sound sensor 1 attached to the user's body surface (for example, the forehead) is based on the impulse-like sound emitted by the user. A bone conduction vibration on the body surface is detected, and a charge signal corresponding to the bone conduction vibration is generated. The charge signal is converted into an electric signal u (t) by the amplifier 21 and input to the transfer function arithmetic processing unit 28 via the switch 24. On the other hand, the microphone 25 detects an impulse-like sound (air conduction sound) emitted by the user, converts it into an electrical signal, and sends the electrical signal to the amplification unit 26. The electric signal sent to the amplifying unit 26 is amplified and then input to the transfer function arithmetic processing unit 28 via the switch 24. After this process is completed, the process proceeds to step S3.

  In step S3, the transfer function calculation processing unit 28, from the input electrical signal u (t) to the input of the impulse sound, from the user's vocal organ to the position where the PVDF bone conduction sensor 1 on the body surface is attached. Calculate the transfer function G (s). After this process ends, the process proceeds to step S4.

In step S 4, the transfer function calculation processing unit 28 calculates an inverse transfer function G (s) ⁻¹ corresponding to the transfer function G (s) calculated in the process of step S 3 and temporarily stores it in the volatile memory 31. . After this process ends, the process proceeds to step S5.

In step S <b> 5, the transfer function calculation processing unit 28 determines whether or not it is necessary to perform correction processing on the inverse transfer function G (s) ⁻¹ . Since this determination is made based on the presence or absence of a correction signal calculated based on a comparison result (error) between a function Y (s) and a function M (s) described later, it is determined that there is no correction at an initial stage, and the process proceeds to step S7. move on. On the other hand, when it determines with correction | amendment, it progresses to step S6. Processing in the case of correction will be described later.

In step S7, the signal processing unit 22 performs Laplace transform on the input electrical signal u (t) to obtain a function U (s) as shown in FIG. 11A, for example. Then, the inverse transfer function G (s) ⁻¹ is convolved with the function U (s) to obtain a function Y (s) of the output audio signal as shown in FIG. 11B, for example. Y (s) is sent to the comparison calculation unit 34. In parallel with this, the transfer function calculation processing unit 28 performs Laplace transform on the input electric signal m (t) to obtain a function M (s) as shown in FIG. (S) is sent to the comparison calculation unit 34. After this process ends, the process proceeds to step S8.

In step S <b> 8, the comparison calculation unit 34 compares the function Y (s) of the output audio signal routed through the PVDF bone conduction sound sensor 1 with the function M (s) of the output signal routed through the microphone 25. By this comparison processing, the inverse transfer function G (s) ⁻¹ used in the signal processing unit 22 is determined depending on whether or not there is a difference between the bone conduction sound detected by the PVDF bone conduction sound sensor 1 and the air conduction sound detected by the microphone 25. It is determined whether or not it is accurate, that is, whether or not the user feels as if he / she is listening to air conduction sound instead of bone conduction sound. After this process ends, the process proceeds to step S9.

  In step S9, when the comparison calculation unit 34 determines that there is an error as a result of the comparison in step S8, the process proceeds to step S10. On the other hand, if it is determined that there is no error in the comparison result between the function Y (s) and the function M (s), the process proceeds to step S11. Here, the presence or absence of an actual error is used as a criterion for determination. However, if the obtained error is within a predetermined threshold range stored in advance in the nonvolatile memory 31 or the like, it is determined that there is no error. It may be.

In step S 10, the comparison calculation unit 34 calculates the correction amount ΔG (s) ⁻¹ of the inverse transfer function G (s) ⁻¹ according to the comparison result of step S 9, that is, the error amount, and sends it to the transfer function calculation processing unit 28. input. After this process is completed, the process proceeds to step S5.

In step S <b> 5, the transfer function calculation processing unit 28 determines whether or not it is necessary to perform correction processing on the inverse transfer function G (s) ⁻¹ obtained previously. Here, since there is an input of a correction signal, it is determined that there is correction, and the process proceeds to step S6.

In step S6, the transfer function calculation processing unit 28 corrects the previously obtained inverse transfer function G (s) ⁻¹ based on the inputted ΔG (s) ⁻¹ . The corrected inverse transfer function G ′ (s) ⁻¹ (= G (s) ⁻¹ + ΔG (s) ⁻¹ ) is temporarily stored in the volatile memory 31 as a new inverse transfer function, and is used for the processing after step S7. used.

After the processing of step S6 is completed, in step S7, the signal processing unit 22 and the transfer function calculation processing unit 28 are each based on the corrected inverse transfer function G ′ (s) ⁻¹ and the PVDF bone conduction sound sensor 1. The function Y (s) of the output audio signal passing through, and the function M (s) of the output signal passing through the microphone 25 are calculated. After this process ends, the process proceeds to step S8.

  In step S <b> 8, the comparison calculation unit 34 compares the function Y (s) of the output audio signal routed through the PVDF bone conduction sound sensor 1 with the function M (s) of the output signal routed through the microphone 25. After this process ends, the process proceeds to step S9.

  In step S9, the comparison calculation unit 34 compares the function Y (s) with the function M (s) again. If it is determined that there is an error as a result of the comparison, the process proceeds to step S10. On the other hand, if it is determined that there is no error by comparing the function Y (s) and the function M (s), the process proceeds to step S11.

  In step 11, the comparison calculation unit 34 stores the inverse transfer function finally determined to have no error in the nonvolatile memory 32. After this process ends, a series of correction mode processes ends.

By such processing in the correction mode (learning function), the inverse transfer function G (s) ⁻¹ used in the signal processing unit 22 can be improved to always accurate and highly reproducible original sound. Thereby, the user can obtain a feeling as if he / she is listening to the air conduction sound instead of the bone conduction sound with respect to the voice captured using the PVDF bone conduction sound sensor 1.

  In the example shown in FIG. 11, when the correction mode is selected, the process is such that the inverse transfer function is obtained one by one (steps S <b> 2 to S <b> 4), but stored in advance in the nonvolatile memory 32 during the registration mode. The inverse transfer function may be used for subsequent processing.

FIG. 12 is a diagram showing a usage example of the PVDF bone conduction sound sensor according to another embodiment of the present invention.
In this embodiment, two PVDF films 11A and 11B are used for the PVDF bone conduction sound sensor, and the detection sensitivity is improved by superimposing them on an arbitrary part of the user's body surface, for example, the forehead. Is. In this case, the vibration detection direction of the PVDF film 11A and the vibration detection direction of the PVDF film 11B are fixed orthogonally. The bone conduction vibration detected in the forehead is also detected from the vertical direction although the horizontal direction of the forehead is dominant. Therefore, by doing in this way, not only the vibration in the horizontal direction V2 of the forehead but also the vibration in the vertical direction V1 can be detected, so that more sensitive and highly accurate detection of the bone conduction vibration can be realized.

  As described above, according to the microphone device of the present invention, the PVDF film used for the PVDF bone conduction sound sensor for detecting the bone conduction vibration need only be brought into close contact with the body surface (skin) of the user, and the strong pressing force can be obtained. Is unnecessary. Therefore, the PVDF film itself can be processed into a soft thin shape with a soft touch. For this reason, it is easy to wear on the user's body surface and can be worn for a long time.

  Moreover, since the detection direction of the biological vibration (bone conduction vibration) of the PVDF film used for the PVDF bone conduction sound sensor is a direction along the body surface, the influence of the sound pressure due to external noise (applied in the vertical direction to the body surface) is affected. Hard to receive and strong in noisy environments.

  Further, since one surface of the PVDF film of the PVDF bone conduction sound sensor 1 is brought into close contact with the skin, there is no influence of resonance derived from the sensor structure, and a flat frequency characteristic can be obtained up to a high frequency range.

  In addition, since bone conduction vibration information is corrected using a reverse transfer function from the user's voice organ to the position where the PVDF bone conduction sensor on the body surface is attached, the reproducibility of the original sound close to the air conduction sound is high. Sound quality is obtained.

  In addition, since the inverse transfer function can be derived from the user's tongue hitting sound as an initial value, it can be used immediately without prior learning. In addition, since the inverse transfer function registered as the initial value can be improved at any time by subsequent learning, the inverse transfer function used in the process of obtaining the output audio signal in the signal processing unit can always be corrected to an accurate one. Reproducibility can be improved.

  Furthermore, from the above, the microphone device of the present invention enables high-quality voice acquisition in a noisy environment, so that communication means (communication device) that can be used in a disaster (for rescue) or in factories, etc. Can be used as Further, by reducing the size of the PVDF bone conduction sound sensor, it can be applied to a mobile phone terminal or the like. Moreover, since the voice that can be acquired has high sound quality, it can be applied to a voice recognition device in a noisy environment. Furthermore, since body vibrations with a wide frequency can be acquired, heart sounds, breathing sounds, eye needles, etc. can be measured over a long period of time. Application to home health monitoring devices used is also conceivable.

  FIG. 13 shows an example of measurement data when a pulse is measured with a PVDF bone conduction sound sensor attached to the neck (throat). The horizontal axis represents time (t), and the vertical axis represents voltage (V). FIG. 14 is an example of measurement data when a PVDF bone conduction sound sensor is attached to the eyelid and the eye needle, that is, the movement of the eye is measured. The horizontal axis represents time (t), and the vertical axis represents voltage (V). Thus, good measurement data can be obtained by detecting fine vibrations.

  In the above-described embodiment, the sensor for detecting the bone conduction sound is configured using the PVDF film. However, the present invention is not limited to this example as long as it is a plate-like piezoelectric element having a piezoelectric effect.

Furthermore, although the above-described embodiment has been described on the assumption that there is only one user, the microphone device of the present invention may be shared by a plurality of users. In this case, for each user, a transfer function from the voice organ to the position where the PVDF bone conduction sensor on the body surface is attached and its inverse transfer function are calculated, and the inverse transfer function is used as the user ID (identification information) and password. Are stored in a non-volatile memory or a recording medium. Each time the use of the microphone device is started, the control calculation unit identifies the user by an ID, a password, etc., reads the corresponding inverse transfer function from the nonvolatile memory, and performs signal processing in the signal processing unit or the like. To use.
In this case, if the inverse transfer function is registered for each user in advance, it is not necessary to measure the inverse transfer function again when it is used later, and the operation is simplified and the usability is improved. In addition, the user can obtain a sound processed by using his / her reverse transfer function without using another person's reverse transfer function, so that a sound with no sense of incongruity can be obtained.

It is a figure which shows the usage example of the PVDF bone-conduction sound sensor which concerns on one embodiment of this invention. It is a perspective view which shows schematic structure of the PVDF bone-conduction sound sensor shown in FIG. It is a schematic sectional drawing (XX arrow) of the PVDF bone-conduction sound sensor shown in FIG. It is a figure which shows the other usage example of the PVDF bone-conduction sound sensor which concerns on one embodiment of this invention. It is a figure which shows the example of the short-time FFT result of the signal obtained from the PVDF bone-conduction sound sensor under white noise. It is a figure which shows the example of the short-time FFT result of the audio | voice signal measured under white noise, (a) is a microphone and (b) is a PVDF bone-conduction sound sensor. It is a figure which shows the example of the short time FFT result of the audio | voice signal measured under the construction site noise, (a) is a microphone, (b) is a PVDF bone-conduction sound sensor. It is a figure which shows the example of the waveform of the audio | voice signal measured under the construction site noise of FIG. 7, (a) is a microphone, (b) is a PVDF bone-conduction sound sensor. 1 is a block diagram showing a schematic configuration of a microphone device to which a PVDF bone conduction sensor according to an embodiment of the present invention is applied. FIG. It is a flowchart which shows the correction process of the reverse transfer function by the microphone apparatus which concerns on one embodiment of this invention. It is a figure which shows the example of the waveform of the audio | voice signal in each block of the microphone apparatus shown in FIG. It is a figure which shows the usage example of the PVDF bone-conduction sound sensor which concerns on other embodiment of this invention. It is a figure which shows the measurement data example of the pulse obtained by applying the microphone apparatus of this invention. It is a figure which shows the measurement data example of the eye needle obtained by applying the microphone apparatus of this invention.

Explanation of symbols

  DESCRIPTION OF SYMBOLS 1 ... PVDF bone-conduction sound sensor, 2 ... Fixing means 11, 11A, 11B ... PVDF film, 12 ... Buffer material, 13 ... Insulating plate, 14 ... Electrode, 15 ... Lead wire, 20 ... Microphone apparatus, 21 ... Amplification part , 22 ... signal processing unit, 23 ... output unit, 24, 27 ... switch, 25 ... microphone, 26 ... amplification unit, 28 ... transfer function arithmetic processing unit, 30 ... program memory, 31 ... volatile memory, 32 ... non-volatile Memory, 33... Control calculation unit, 34... Comparison calculation unit, 35.

Claims (5)

A detection unit that is a piezoelectric element that is plate-shaped and detects vibrations of a living body surface that is in close contact with one surface of the plate-shaped and converts it into an electrical signal;
The detection unit on the surface of the living body is attached from the sounding organ of the living body based on an electrical signal corresponding to the vibration of the surface of the living body due to the impulse-like sound uttered from the generating organ of the living body output from the detecting unit. A calculation unit for calculating a transfer function of the living body up to a given position and its inverse transfer function;
A signal processing unit that performs a filter process by applying a filter that performs the calculation using the inverse transfer function calculated by the calculation unit to the electrical signal output from the detection unit;
An output unit for outputting the electrical signal filtered by the signal processing unit to the outside;
A sound collection unit that collects the air conduction sound uttered from the living body's vocal organs and propagates through the space, and converts it into an electrical signal;
The electric signal to which the filter is applied in the signal processing unit is compared with the electric signal output from the sound collection unit, and the correction amount of the inverse transfer function calculated by the calculation unit based on the comparison result is obtained. A comparison operation unit for calculating,
The microphone unit corrects the inverse transfer function based on the correction amount calculated by the comparison calculation unit. The microphone device according to claim 1 , wherein the impulse-like sound uttered from the voicing organ of the living body is a user's tongue hitting sound. A cushioning material provided on the opposite side of the surface close to the living body surface of the plate-like detection unit,
Said plate-shaped detector, microphone apparatus according to claim 1 or 2, characterized in that it is held between the one surface of said living body surface the cushioning material. The microphone device according to claim 3 , wherein the cushioning material is formed in a spongy shape. The microphone device according to any one of claims 1 to 4 , wherein the piezoelectric element of the detection unit has a plate shape made of polyvinylidene fluoride.