JP2006126558A - Voice speaker authentication system - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a voice speaker authentication system which prevents "pretence" by synthesized voice using recorded voice or voice synthesis technology, and is hardly affected by ambient noise. <P>SOLUTION: When releasing door lock, an utterer 100 who is an authentication registrant is mounted with a NAM microphone 101, and utters a keyword for authentication in a non-audible murmur note (NAM note) or a usual note. The NAM note or the usual note is picked up by the NAM microphone 101, is amplified by a mic-amplifier 103 and digitized by an A-D converter 104, thereafter, input to a voice authentication part 106. The voice authentication part 106 calculates a degree of similarity between a voice pattern which has been created and registered based on the NAM note or the usual note pre-uttered for registration and has been stored in a registration pattern storage part 107, and the NAM note or the usual note uttered in authentication, and compares the degree of similarity with a preset threshold value to judge whether or not the utterer 100 is authorized to enter or exit from a room. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

  The present invention relates to a voice speaker authentication system for identifying a speaker by using voice, and in particular, mounted on the skin surface on the sternocleidomastoid muscle of the lower rear part of the auricle, directly below the mastoid process of the skull, Non-Audible Murmur (Non-Audible Murmur), which is a vibration sound transmitted through the soft tissue of the body that is not audible from the outside and is not accompanied by the regular vibration of the vocal cords, and is tuned by the resonance filter characteristic change accompanying the movement of The present invention relates to a voice speaker authentication system that uses a voice collected from a microphone (NAM microphone) that collects and inputs a body conduction voice that conducts through soft tissue in the body such as NAM) or normal voice.

Conventionally, a magnetic card, an IC card, a password, and the like have been used as a method for confirming the identity of the person in entrance / exit management or an ATM of a financial institution. However, there is a risk of loss, theft and forgery of magnetic cards and IC cards, and passwords may be forgotten or stolen.
For this reason, in recent years, biometric authentication such as fingerprint authentication, iris authentication, and voice authentication using the physical characteristics of the person has been proposed. Of these, voice authentication is a method that uses individual differences in voice. Regarding the algorithm, for example, as described in Non-Patent Document 1, first, a keyword is uttered and a voice pattern is registered. The same keyword is uttered, the similarity with the registered pattern is calculated, and compared with a preset threshold value to determine whether or not the user is the person. Voice authentication can be realized at low cost without the need for special input devices such as those used for fingerprints and irises, and has the advantage of being able to authenticate from a remote location using a telephone. Etc. are introduced.

In order to make identity verification systems based on voice authentication more widespread, there are issues such as countermeasures against fraud (hereinafter referred to as “spoofing”), such as voice recordings using tape recorders and synthesized voices using voice synthesis technology. For example, the false authentication due to noise can be reduced.
The following countermeasures have been proposed for this “spoofing” problem. That is, if multiple keywords are registered in advance and there is no voice input within the time when the keyword is randomly specified by the system, even if the similarity to the registered pattern is large, it is judged as synthesized voice or recorded voice If the keyword is spoken multiple times during authentication and the similarity between the multiple voices is close to perfect match, the input voice is not a synthesized voice or recorded voice. A coping method (for example, see Patent Document 2) such as determining that the sound is natural is proposed.
Japanese Patent Application Laid-Open No. 61-272798 JP 2001-265387 A Yoichi Seto “Biometric Authentication Technology” Kyoritsu Publishing Co., Ltd. May 2002 pp. 64-68

However, with recent advances in speech synthesis technology, it is possible to easily generate the speech of the person to be authenticated in a short time, and to make the synthesized speech fluctuate by adjusting the speech synthesis parameters. There is a problem that it is no longer possible to cope with “spoofing”.
For the problem of ambient noise, it is conceivable to use a directional microphone, but there is no drastic countermeasure at present.
The present invention has been made in order to solve the above-mentioned problems, and its purpose is to prevent “spoofing” by a recorded voice by a tape recorder or the like and a synthesized voice using a voice synthesis technique, and to prevent the influence of ambient noise. It is to provide a voice speaker authentication system that is difficult to receive.

  According to a first aspect of the present invention, there is provided a voice speaker authentication system in which data relating to body conduction speech for authentication is stored in advance, and body conduction speech for inputting body conduction speech of the person to be authenticated from the skin surface. An authentication unit that performs authentication by collating data related to body conduction speech input by the body conduction speech input unit and data related to body conduction speech stored in the storage unit; The external device is controlled according to the authentication result of the means. With this configuration, it is possible to prevent “spoofing” due to a voice recorded by a tape recorder or the like and a synthesized voice using a voice synthesis technique, and is less susceptible to ambient noise.

The voice speaker authentication system according to claim 2 of the present invention is characterized in that, in claim 1, the in-vivo conduction voice input means is a NAM (Non-Audible Murmur) microphone. In this way, using an NAM microphone makes it possible to easily input inaudible tweets.
A voice speaker authentication system according to claim 3 of the present invention is characterized in that, in claim 1 or 2, the in-body conduction speech is a non-audible murmur. By making a non-audible murmur sound as an authentication target, even when the utterance of the authentication target person is not in a normal state due to a cold or the like, it can be properly authenticated.

The voice speaker authentication system according to claim 4 of the present invention is characterized in that, in claim 2 or 3, the NAM microphone is configured to be detachable from the own system. If the NAM microphone is configured to be detachable from the system as described above, the NAM microphone can be carried and the convenience of the system is improved.
A voice speaker authentication system according to claim 5 of the present invention is the voice speaker authentication system according to any one of claims 1 to 4, wherein data relating to the in-body conduction speech input by the in-body conduction speech input means is authenticated by a wireless communication method. Further comprising wireless communication means for communicating to the means. With this configuration, a part of the system configuration can be carried and the convenience of the system is improved.

  The voice speaker authentication system according to claim 6 of the present invention is the method for inputting identification information for identifying the authentication target person given in advance to the authentication target person in any one of claims 1 to 5. The authentication means is inputted by the body conduction voice input means and data corresponding to the identification information inputted by the ID input means among the data stored in the storage means. It is characterized by collating with the data concerning the conduction sound in the body. If comprised in this way, it can respond to a some authentication subject.

  According to a seventh aspect of the present invention, there is provided the voice speaker authentication system according to the sixth aspect, wherein the ID input means inputs the identification information by reading the identification information stored in a non-contact ID tag. Features. If a non-contact ID tag such as RFID (Radio Frequency Identification) is used, identification information for identifying the person to be authenticated can be easily input.

  The voice speaker authentication system according to claim 8 of the present invention is the voice speaker authentication system according to any one of claims 1 to 7, wherein the authentication means includes the data related to the body conduction speech input by the body conduction speech input means and the storage. The similarity with the data related to the in-body conduction speech stored in the means is compared with a predetermined threshold value. If comprised in this way, the data regarding the in-vivo conduction sound, such as a non-audible murmur sound, and the data regarding the in-body conduction sound, such as a non-audible muzzle sound, which are input in advance can be easily collated.

The voice speaker authentication system according to claim 9 of the present invention is the voice speaker authentication system according to any one of claims 1 to 8, wherein the external device controls locking of the door by an electrical signal corresponding to the authentication result. And If comprised in this way, locking of a door can be controlled.
The voice speaker authentication system according to claim 10 of the present invention is the voice speaker authentication system according to any one of claims 1 to 8, wherein the external device controls whether to log in to the network based on data corresponding to the authentication result. Features. With this configuration, it is possible to control whether to log in to the network.
The voice speaker authentication system according to an eleventh aspect of the present invention is the voice speaker authentication system according to any one of the first to eighth aspects, wherein the external device controls whether or not to access a database based on data corresponding to the authentication result. Features. With this configuration, it is possible to control whether or not access to the database is possible.

According to the present invention, resistance to “spoofing” is dramatically improved, and a voice authentication system with high security can be constructed. Furthermore, if NAM voice is used for authentication, voice authentication can be performed without a third party being asked for an authentication keyword, thereby further improving the security of the system.
In addition, since the NAM microphone is less affected by environmental noise, it is possible to realize a voice authentication system having higher authentication performance.

Hereinafter, embodiments of the present invention will be described with reference to the drawings. In the drawings referred to in the following description, the same parts as those in the other drawings are denoted by the same reference numerals.
First, inaudible tweet (NAM) is a vibration sound that is tuned by the resonance filter characteristic change accompanying the movement of the vocal organs, and that is not accompanied by the regular fluctuation of the vocal cords and that is transmitted from the outside to the soft tissue of the body that is inaudible. It is.

(principle)
Since the meat conduction sound collected by the NAM microphone proposed in the pamphlet of International Publication No. WO 2004/021738 passes through a substance having a plurality of different acoustic impedances in the process, the signal of the signal in the high frequency band is particularly compared with the air conduction. Attenuation is large. Further, since the location of the sound source changes depending on the phoneme and prosody, the path of the meat conduction is not uniform and the attenuation characteristics are also different. In the present invention, it is considered to prevent “spoofing” by utilizing the difference between the sound conduction characteristic and the air conduction characteristic of sound.

  First, the inventor made a prototype of a NAM microphone as shown in FIGS. FIG. 2 is a side sectional view of the NAM microphone, and FIG. 3 is a front view of the NAM microphone. The NAM microphone 101 shown in both figures includes a contact portion 301a, a frame 301b, an external noise soundproof space 301c, a reflecting plate 301d, and a condenser microphone 301e. The condenser microphone 301e has a diaphragm electrode 301f and a lead wire 301g. The contact portion 301a is made of silicone rubber, which is a biocompatible material having an acoustic impedance close to that of human soft tissue so that acoustic impedance mismatch with the skin does not occur.

FIG. 4A is a diagram showing a spectrum of a signal obtained from the NAM microphone 101 placed at the mouth of air conduction sound generated by uttering a sentence of a certain content, and FIG. It is a figure which shows the spectrum of the signal which extract | collected the meat conduction sound produced when the same speaker uttered the sentence of the same content from the NAM microphone 101 with which the body surface was mounted | worn.
Referring to FIG. 5A, the air conduction sound includes a signal of 4 kHz or more. On the other hand, referring to FIG. 5B, it can be seen that the signal of 4 kHz or more is attenuated in the meat conduction sound.

  FIG. 5 (a) is a diagram in which an air conduction sound produced by each speaker uttering “A” is sampled from the NAM microphone 101 placed at the mouth and subjected to octave analysis, and FIG. 5 (b) is illustrated. FIG. 4 is a diagram obtained by performing octave analysis on the meat conduction sound generated by the same speaker uttering “A” and collected from the same NAM microphone 101 mounted on the body surface. In both figures, the center frequencies are 80 Hz, 100 Hz, 125 Hz, 160 Hz, 200 Hz, 250 Hz, 315 Hz, 400 Hz, 500 Hz, 650 Hz, 800 Hz, 1000 Hz, 1250 Hz, 1600 Hz, 2000 Hz, 2500 Hz, 3150 Hz, 4000 Hz, and 5000 Hz, respectively, from the left. .

  Similarly, FIG. 6 (a) and FIG. 6 (b) say “Yes”, FIG. 7 (a) and FIG. 7 (b) say “Yes”, FIG. 8 (a) and FIG. 8 (b) is when “e” is uttered, and FIGS. 9 (a) and 9 (b) are when “o” is uttered. Referring to these figures, for any vowel, the frequency pattern of meat conduction sound collected with the NAM microphone 101 attached to the body surface and the frequency pattern of air conduction sound collected from the NAM microphone 101 placed at the mouth. In addition, it can be seen that the power ratio of air conduction sound and meat conduction sound for each frequency band in the high frequency range is different for “A”, “I”, “U”, “E”, “O”.

Therefore, even if the high band of air conduction sound is simply attenuated by a filter or the like, it does not match the frequency pattern of meat conduction sound. In normal “spoofing”, the air conduction sound of the “spoofing” subject is recorded / analyzed, and the air conduction sound is reproduced / synthesized and input from a microphone such as a speaker. Therefore, if an authentication pattern for registration is created based on the meat conduction sound collected with the NAM microphone attached to the body surface, the recorded voice or synthesized voice of the person impersonated by the above method can be used by a third party. Even if the voice is input to the NAM microphone, the pattern does not match and “spoofing” can be prevented. In addition, even if an attempt is made to generate a meat conduction sound based on the subject's air conduction sound and input it to the NAM microphone, the target meat conduction sound is not known by a third party, which is very difficult in reality. .
Furthermore, since it is not used everyday, training is required until it can be uttered stably. However, if the authentication pattern for registration is created from the NAM uttered speech, the keyword itself will not be known to third parties, and security will be improved. Further improve.

Next, the environmental noise resistance characteristics of the NAM microphone 101 will be described. First, as shown in FIG. 10 (a), the NAM microphone 101 in which air conduction sound and flesh conduction sound when a certain speaker utters “a” is attached to the skin surface directly below the normal microphone MC and the mastoid process. And were collected respectively. Next, as shown in FIG. 10B, the above-mentioned “a” of the speaker voice normally collected by the microphone MC from the speaker installed at the speaker position is collected in FIG. The gain was adjusted so that the volume was the same as that of the speaker voice, and the air conduction sound was collected from the NAM microphone 101 mounted on the skin surface immediately below the speaker's mastoid next to the speaker SP.
The meat conduction sound collected from the NAM microphone 101 is shown in FIG. 11 (a), and the air conduction sound collected from the NAM microphone 101 is shown in FIG. 11 (b). Referring to FIGS. 4A and 4B, it can be seen that the air conduction sound is attenuated by 30 dB or more compared to the meat conduction sound.

  FIG. 1 is a diagram showing a configuration example of a door lock unlocking system based on voice authentication according to the present invention. The door lock unlocking system shown in FIG. 1 includes a NAM microphone 101, a voice preprocessing unit 102, a voice authentication system 105, a door lock control system 108, and a door 109. Furthermore, the voice preprocessing unit 102 includes a microphone amplifier 103 and an AD converter 104, and the voice authentication system 105 includes a voice authentication unit 106 and a registered pattern storage unit 107.

  When unlocking the door lock, the utterer 100 who is an authentication registrant wears the NAM microphone 101 on the skin surface on the sternocleidomastoid muscle in the lower rear part of the auricle, directly below the mastoid process of the skull, The authentication keyword is uttered with a non-audible murmur (NAM sound) or normal sound. The NAM sound or normal sound is collected by the NAM microphone 101 and input to the microphone amplifier 103. The NAM sound or normal sound amplified by the microphone amplifier 103 is digitized by the AD converter 104 and then input to the voice authentication unit 106. The voice authentication unit 106 is created and registered based on the NAM sound or normal sound uttered in advance for registration, and is stored in the registered pattern storage unit 107, and the NAM sound or normal sound uttered at the time of authentication. Is compared with a threshold value set in advance, and it is determined whether or not the speaker 100 has authority to enter and exit the room. That is, the voice authentication unit 106 performs authentication by collating the input data related to the inaudible murmur and the data related to the inaudible murmur stored in the registered pattern storage unit 107.

The determination result obtained by the voice authentication unit 106 is transmitted to the door lock control system 108. When the door lock control system 108 determines that the user has authority to enter and exit according to the determination result, the door lock of the door 109 is fixed. Unlock for hours.
Note that the speaker 100, who is the subject of authentication, may utter using either the NAM utterance or the normal utterance, but it is necessary to use the same utterance method during registration and during authentication. However, it is also possible to prepare registration patterns for both NAM utterances and normal utterances at the time of registration, and to select an utterance method according to the surrounding situation at the time of authentication.

  FIG. 12 is a diagram showing another configuration example of the door lock unlocking system by voice authentication according to the present invention. In the configuration of FIG. 12, a radio transmission unit 201 and a radio reception unit 204 are newly added to the configuration of FIG. 1, and the radio transmission unit 201 includes a digital modulation unit 202 and a transmission antenna 203. A receiving antenna 205 and a digital demodulator 206 are included.

When unlocking the door lock, the utterer 100 who is an authentication registrant wears the NAM microphone 101 on the skin surface on the sternocleidomastoid muscle in the lower rear part of the auricle, directly below the mastoid process of the skull, The authentication keyword is uttered with a non-audible murmur (NAM sound) or normal sound. The NAM sound or normal sound is collected by the NAM microphone 101 and input to the microphone amplifier 103. The NAM sound or normal sound amplified by the microphone amplifier 103 is digitized by the AD converter 104, then digitally modulated by the digital modulation unit 202, and sent to the digital demodulation unit 206 via the transmission antenna 203 and the reception antenna 205. It is done. The digital demodulation unit 206 extracts the original digital audio data from the received signal and inputs it to the audio authentication unit 106.
Subsequent operations are the same as those in the first embodiment, and will be omitted.

  FIG. 13 is a diagram showing still another configuration example of the door lock unlocking system by voice authentication according to the present invention. The configuration shown in the figure is different from that in the first embodiment (FIG. 1) in that a personal ID input unit 1402 is added to the voice authentication system 1401, and the voice authentication unit 1403 receives digitized voice data from the AD converter 104. In addition, the personal ID information from the personal ID input unit 1402 is also input.

  When it is desired to unlock the door lock, the speaker 100 who is an authentication registrant inputs a personal ID from the personal ID input unit 1402 using an ID card or a numeric keypad, and directly below the mastoid process of the skull at the lower rear part of the auricle. The NAM microphone 101 is attached to the skin surface on the sternocleidomastoid muscle, and the authentication keyword is uttered by an inaudible murmur (NAM sound) or a normal sound. The NAM sound or normal sound is collected by the NAM microphone 101 and input to the microphone amplifier 103. The NAM sound or normal sound amplified by the microphone amplifier 103 is digitized by the AD converter 104 and then input to the voice authentication unit 1403.

The voice authentication unit 1403 is created and registered based on the NAM sound or normal sound uttered for registration corresponding to the personal ID input from the personal ID input unit 1402 and stored in the registration pattern storage unit 107. The similarity between the voice pattern and the NAM sound or normal sound uttered at the time of authentication is calculated and compared with a preset threshold value to determine whether or not the speaker 100 has authority to enter and exit the room. .
Subsequent operations are the same as those in the first embodiment, and will be omitted.

  FIG. 14 is a diagram showing a configuration example of a network login management system by voice authentication according to the present invention. Referring to FIG. 2, the network login management system includes a NAM microphone 101, a voice preprocessing unit 102, a voice authentication system 1401, a network login management unit 1501, and a network 1502. Furthermore, the audio preprocessing unit 102 includes a microphone amplifier 103 and an AD converter 104. The voice authentication system 1401 includes a personal ID input unit 1402, a voice authentication unit 1403, and a registered pattern storage unit 107.

  When logging in to the network 1502, the speaker 100 who is an authentication registrant inputs a personal ID from the personal ID input unit 1402 using an ID card or a numeric keypad, and directly below the mastoid process of the skull, The NAM microphone 101 is attached to the skin surface on the sternocleidomastoid muscle, and the authentication keyword is uttered with an inaudible murmur (NAM sound) or a normal sound. The NAM sound or normal sound is collected by the NAM microphone 101 and input to the microphone amplifier 103. The NAM sound or normal sound amplified by the microphone amplifier 103 is digitized by the AD converter 104 and then input to the voice authentication unit 1403.

The voice authentication unit 1403 is created and registered based on the NAM sound or normal sound uttered for registration corresponding to the personal ID input from the personal ID input unit 1402 and stored in the registration pattern storage unit 107. The similarity between the voice pattern and the NAM sound or normal sound uttered at the time of authentication is calculated and compared with a preset threshold value to determine whether or not the speaker 100 has authority to enter and exit the room. .
The determination result obtained by the voice authentication unit 1403 is transmitted to the network login management unit 1501, and the network login management unit 1501 permits login to the network 1502 when it is determined that the user has login authority according to the determination result. To do.

  FIG. 15 is a diagram showing a configuration example of a database access management system based on voice authentication according to the present invention. Referring to the figure, the database access management system includes a NAM microphone 101, a voice preprocessing unit 102, a voice authentication system 1401, a database access management unit 1601, and a database 1602. Further, the voice preprocessing unit 102 includes a microphone amplifier 103 and an AD converter 104, and the voice authentication system 1401 includes a personal ID input unit 1402, a voice authentication unit 1403, and a registered pattern storage unit 107.

  To access the database 1602, the speaker 100 who is an authentication registrant inputs a personal ID from the personal ID input unit 1402 using an ID card or a numeric keypad, and directly below the mastoid of the skull at the lower rear part of the auricle. The NAM microphone 101 is attached to the skin surface on the sternocleidomastoid muscle, and the authentication keyword is uttered with an inaudible murmur (NAM sound) or a normal sound. The NAM sound or normal sound is collected by the NAM microphone 101 and input to the microphone amplifier 103. The NAM sound or normal sound amplified by the microphone amplifier 103 is digitized by the AD converter 104 and then input to the voice authentication unit 1403. The voice authentication unit 1403 is created and registered based on the NAM sound or normal sound uttered for registration corresponding to the personal ID input from the personal ID input unit 1402 and stored in the registration pattern storage unit 107. The similarity between the voice pattern and the NAM sound or normal sound uttered at the time of authentication is calculated and compared with a preset threshold value to determine whether or not the speaker 100 has authority to enter and exit the room. . The determination result obtained by the voice authentication unit 1403 is transmitted to the database access management unit 1601, and the database access management unit 1601 permits access to the database 1602 when it is determined that the user has access authority according to the determination result. To do.

  FIG. 16 is a diagram showing still another configuration example of the door lock unlocking system by voice authentication according to the present invention. Referring to the figure, the door lock unlocking system includes an NAM microphone 1701 with an RFID tag, a microphone input terminal 1702, a voice preprocessing unit 102, a voice authentication system 1703, a door lock control system 108, a door 109, It is composed of Furthermore, the audio preprocessing unit 102 includes a microphone amplifier 103 and an AD converter 104. The voice authentication system 1703 includes an RFID tag reader / writer 1704, a voice authentication unit 1705, and a registered pattern storage unit 107.

Each speaker 100 who is an authentication registrant has an individual NAM microphone 1701 with an RFID tag, and personal ID information is written in the RFID tag in advance.
A configuration example of the NAM microphone 1701 with an RFID tag will be described with reference to FIGS. FIG. 17 is a side sectional view of the NAM microphone 1701 with an FID tag, and FIG. 18 is a front view of the NAM microphone 1701 with an FID tag. The configuration of the NAM microphone 1701 shown in FIGS. 17 and 18 is different from the configuration of the NAM microphone 101 shown in FIGS. 2 and 3 in that the RFID tag 1801 in which personal ID information is written in advance. A plug 1802 corresponding to the microphone input terminal 1702 is added.

Since the identification information is input by reading the identification information stored in a non-contact ID tag such as an RFID, the identification information for identifying the person to be authenticated can be easily input.
Since the plug 1802 is added, the NAM microphone 1701 is configured to be detachable from the system. If comprised in this way, a NAM microphone can be carried and the convenience of a system will improve.

When it is desired to unlock the door lock, the speaker 100 who is an authentication registrant first brings the RFID tag-attached NAM microphone 1701 close to the RFID tag reader / writer 1704 for each individual. The RFID tag reader / writer 1704 reads the personal ID from the RFID tag 1801 of the RFID tag-attached NAM microphone 1701.
Next, the speaker 100 who is an authentication registrant inserts the plug 1802 of the RFID tag-attached NAM microphone 1701 into the microphone input terminal 1702, and then the thoracic milk in the lower rear part of the auricle just below the mastoid process of the skull. A NAM microphone 1701 with an RFID tag is attached to the skin surface above the gluteal muscle, and the authentication keyword is uttered with an inaudible murmur (NAM sound) or a normal sound.

  The NAM sound or normal sound is collected by the RFID tag-attached NAM microphone 1701 and input to the microphone amplifier 103. The NAM sound or normal sound amplified by the microphone amplifier 103 is digitized by the AD converter 104 and then input to the voice authentication unit 1705. The voice authentication unit 1705 is created and registered based on the NAM sound or normal sound uttered for registration corresponding to the personal ID read by the RFID tag reader / writer 1704, and stored in the registration pattern storage unit 107. The similarity between the pattern and the NAM sound or normal sound uttered at the time of authentication is calculated, and compared with a preset threshold value, it is determined whether or not the speaker 100 has the right to enter and leave the room.

Subsequent operations are the same as those in the first embodiment, and will be omitted.
For each individual, possessing and using a NAM microphone with RFID tag 1701 that combines a small and inexpensive NAM microphone and an RFID tag makes it possible to enter a personal ID without any special operation. A highly reliable system can be realized.
In any of the first to fifth embodiments, if a plug corresponding to the microphone input terminal is added, the NAM microphone can be configured to be detachable from each system, and the NAM microphone can be carried around. Convenience is improved.

  According to the present invention, a voice authentication system with high security and high authentication performance can be realized. Therefore, it can be widely used as a personal authentication system in security systems and financial systems other than entrance management and ATMs. It becomes possible.

It is a figure showing composition of a door lock unlocking system concerning an embodiment of the invention. It is side surface sectional drawing which shows the structural example of a NAM microphone. It is a front view which shows the structural example of a NAM microphone. It is a figure which shows the spectrum of the signal extract | collected from the NAM microphone. It is an octave analysis diagram of air conduction sound and meat conduction sound of utterance "A". It is an octave analysis diagram of air conduction sound and meat conduction sound of utterance “I”. It is an octave analysis diagram of air conduction sound and meat conduction sound of utterance “U”. It is an octave analysis diagram of air conduction sound and meat conduction sound of utterance “E”. It is an octave analysis diagram of air conduction sound and meat conduction sound of utterance "O". It is an evaluation block diagram for environmental noise resistance of a NAM microphone. It is a wave form diagram of the meat conduction sound and air conduction sound which were extract | collected from the NAM microphone. It is a figure which shows the structure of another door lock unlocking system concerning this invention. It is a figure which shows the structure of another door lock unlocking system concerning this invention. It is a figure which shows the structure of the network login management system concerning this invention. It is a figure which shows the structure of the database access management system concerning this invention. It is a figure which shows the structure of another door lock unlocking system concerning this invention. It is side surface sectional drawing which shows the structural example of the NAM microphone with an RFID tag. It is a front view which shows the structural example of the NAM microphone with an RFID tag.

Explanation of symbols

DESCRIPTION OF SYMBOLS 100 Speaker 101 NAM microphone 102 Voice preprocessing part 103 Microphone amplifier 104 AD converter 105 Voice authentication system 106 Voice authentication part 107 Registration pattern memory | storage part 108 Door lock control system 109 Door 201 Wireless transmission unit 202 Digital modulation part 203 Transmitting antenna 204 Wireless receiving unit 205 Reception antenna 206 Digital demodulator 301a Contact 301b Frame 301c External noise soundproof space 301d Reflector 301e Capacitor microphone 301f Diaphragm electrode 301g Lead wire 1401 Voice authentication system 1402 Personal ID input part 1403 Voice authentication part 1501 Network login Management unit 1502 Network 1601 Database access management unit 1602 Database 1701 With RFID tag NAM microphone 1702 Microphone input terminal 1703 Voice authentication system 1704 RFID tag reader / writer 1705 Voice authentication unit 1801 RFID tag 1802 Plug MC Normal microphone SP Speaker

Claims (11)

  Data relating to the body conduction voice for authentication is stored in advance, the body conduction voice input means for inputting the body conduction voice of the person to be authenticated from the skin surface, and the body conduction voice input means. Authentication means for performing authentication by comparing data relating to body conduction speech and data relating to body conduction speech stored in the storage means, and to control an external device according to the authentication result of the authentication means A voice speaker authentication system characterized by   2. The voice speaker authentication system according to claim 1, wherein the body conduction voice input means is a NAM (Non-Audible Murmur) microphone.   The voice speaker authentication system according to claim 1, wherein the body conduction voice is a non-audible murmur.   4. The voice speaker authentication system according to claim 2, wherein the NAM microphone is configured to be detachable from its own system.   5. The wireless communication unit according to claim 1, further comprising a wireless communication unit configured to transmit data related to the internal conductive speech input by the internal body speech input unit to the authentication unit by a wireless communication method. The voice speaker authentication system described.   It further includes ID input means for inputting identification information given in advance to the authentication target person and for identifying the authentication target person, wherein the authentication means includes the ID among the data stored in the storage means. The data corresponding to the identification information inputted by the input means is collated with the data related to the body conduction voice inputted by the body conduction voice input means, according to any one of claims 1 to 5. The voice speaker authentication system described.   7. The voice speaker authentication system according to claim 6, wherein the ID input means inputs the identification information by reading the identification information stored in a non-contact ID tag.   The authentication means compares the similarity between the data related to the body conduction speech input by the body conduction speech input means and the data related to the body conduction speech stored in the storage means, and a predetermined threshold value. The voice speaker authentication system according to any one of claims 1 to 7.   The voice speaker authentication system according to any one of claims 1 to 8, wherein the external device controls the locking of the door by an electrical signal corresponding to the authentication result.   The voice speaker authentication system according to any one of claims 1 to 8, wherein the external device controls whether or not to log in to the network based on data corresponding to the authentication result.   The voice speaker authentication system according to any one of claims 1 to 8, wherein the external device controls whether or not to access a database based on data corresponding to the authentication result.