JP2004271561A - Speech input device, speaker identification device using the same, speech input method and speaker identification method using the same, speech input program, speaker identification program, and program recording medium - Google Patents

Abstract

An object of the present invention is to obtain voice data and speaker identification information delimited by an utterance section without imposing a heavy burden on a user.
When a voice input switch is turned on, biological information acquired by a biological information acquisition sensor is input from a biological information input unit and a voice is input from a voice input unit. The biometric information collating unit 4 collates the input biometric information from the biometric information input unit 2 with the registered biometric information registered in the biometric information registration unit 3, and obtains a speaker identification result based on the collation result. Thus, by operating the voice input switch 5 and automatically acquiring the biometric information of the speaking speaker, it is not necessary to individually input the speaker information. Therefore, it is possible to obtain the voice data and the speaker identification information separated by the utterance section without putting a burden on the speaker.
[Selection diagram] Fig. 1

Description

[0001]
TECHNICAL FIELD OF THE INVENTION
The present invention relates to a voice input device for inputting voice, a speaker identification device using the same, a voice input method and a speaker identification method using the same, a voice input program, a speaker identification program, and a program recording medium.
[0002]
[Prior art]
Conventionally, there is a user identification device for identifying a user using fingerprint data or voice data. For example, in a fingerprint authentication device disclosed in Japanese Patent Application Laid-Open No. 2000-165378, a mobile wireless terminal is operated based on an authentication result of fingerprint data read by a fingerprint reading unit provided in a power switch, and is input from a microphone. The mobile radio terminal can communicate based on the authentication result of the voice data.
[0003]
Further, for example, in a mobile phone disclosed in Japanese Patent Application Laid-Open No. H11-146057, user's voice data, image data such as pinna and iris, fingerprint and palm print irregularity data, The mobile phone is enabled to talk based on the identification result.
[0004]
In addition, conventionally, there is a speech recognition device that adapts a speech dictionary to a specific user and registers the speech dictionary when the user speaks in advance. For example, in ViaVoice of IBM Japan, Ltd., a user speaks a large amount of voice to register a speech dictionary after adapting to a specific user, and after registration, the user indicates the registered speech dictionary. By selecting an index, speech is recognized using a speech dictionary adapted to the user.
[0005]
[Problems to be solved by the invention]
However, the above-described conventional user identification device has the following problems with respect to input of audio data. That is, first, in the fingerprint authentication device disclosed in Japanese Patent Application Laid-Open No. 2000-165378, voice during a call is input from a microphone for authentication, but the range of voice data to be authenticated (acquisition timing) is described. It is not mentioned, and it is technically difficult to accurately authenticate the entire voice of the call due to the influence of ambient noise and the like. Further, in the mobile telephone disclosed in Japanese Patent Application Laid-Open No. H11-146057, the security effect is improved by identification based on various user information. However, input of user information is handled individually, and there is a problem that a burden is imposed on the user in inputting.
[0006]
In addition, the above-described conventional speech recognition apparatus has the following problem regarding selection of a registered speech dictionary. In the case of ViaVoice of IBM Japan, Ltd., it is necessary for the user to select an index indicating a registered voice dictionary. However, when there are a plurality of users using the same voice recognition device, When the index is intense, selecting the index every time the user changes places a burden on the user.
[0007]
Therefore, an object of the present invention is to provide a voice input device capable of obtaining voice data and speaker identification information separated by a voice section without imposing a heavy burden on a user, a speaker identification device using the voice input device, and a voice An object of the present invention is to provide an input method, a speaker identification method using the same, a voice input program, a speaker identification program, and a program recording medium.
[0008]
[Means for Solving the Problems]
In order to achieve the above object, in the voice input device of the present invention, a speaker switches a voice input switch to enable voice input by a voice input unit and to enable biological information acquisition by a biological information acquisition unit. Can be Then, voice data is generated by the voice input means based on the voice of the speaker and input to the subsequent stage. Further, the biological information acquired by the biological information acquiring means is input from the biological information input means. Then, the input biometric information and the registered biometric information registered in the biometric information registration means are compared and collated by the biometric information matching means, a speaker associated with the input voice is identified, and a speaker identification result is output.
[0009]
Thus, the biometric information of the speaker speaking by operating the voice input switch is automatically acquired, and the speaker identification is performed. Therefore, it is not necessary to handle the input of the speaker information individually, and the voice data and the speaker identification information separated by the utterance section can be obtained without putting a burden on the speaker. In addition, since voice data is not used for authentication, the speaker is accurately authenticated without being affected by ambient noise. Further, if the speaker identification information is used for selecting a speaker adaptive dictionary used for speech recognition, it becomes possible to automatically select the speaker adaptive dictionary.
[0010]
In the voice input device according to the embodiment, the biometric information includes a fingerprint or a palm print. In this way, speaker identification is easily performed based on speaker information that is relatively easy to acquire at the time of voice input.
[0011]
Further, in the voice input device of one embodiment, the voice input switch is in an on state that enables the voice input and the acquisition of the biological information while being pressed, and the voice input switch is not pressed when the voice input switch is not pressed. It is assumed that the state is an off state in which the acquisition of biological information is to be rejected. Therefore, as in the case of a general switch, the voice input switch is turned on while the voice input switch is kept depressed, so that the voice input and the biological information acquisition can be performed.
[0012]
Further, in the voice input device of one embodiment, when the voice input switch is pressed, the voice input device is turned on to enable the voice input and the acquisition of biological information, and the utterance end determination unit determines that the utterance has ended. Then, the state is automatically set to an off state in which the voice input and the biological information acquisition are denied. In this way, the voice data delimited by the utterance section is automatically obtained without imposing a burden on the speaker.
[0013]
Further, in the voice input device of one embodiment, the biometric information acquisition means is configured to continuously obtain biometric information while the voice input switch is on. Thus, it is possible to input accurate biological information.
[0014]
Further, in the voice input device of one embodiment, the biometric information acquisition means is configured to obtain the biometric information only once when the voice input switch is turned on. Thus, the biological information input process is performed quickly.
[0015]
Also, in the speaker identification device according to the present invention, the speaker enables the voice input switch of the voice input device to enable voice input by the voice input unit, and also enables biological information acquisition by the biological information acquisition unit. Then, the voice input device inputs voice data based on the voice of the speaker to the subsequent stage, and outputs a speaker identification result related to the input voice based on the input biometric information.
[0016]
Further, voiceprint information is extracted from the voice data by voiceprint information extracting means provided in the voice input means of the voice input device. Then, the input voiceprint information and the registered voiceprint information registered in the voiceprint information registration unit are compared and collated by the voiceprint collation unit, a speaker related to the input voice is identified, and a speaker identification result is output. Then, the result of speaker identification by the biometric information matching means of the voice input device and the result of speaker identification by the voiceprint matching means are integrated by the matching result integration means, and one speaker identification result is output.
[0017]
Thus, the voice input switch is operated to automatically input the biometric information and the voiceprint information of the speaking speaker, thereby performing the speaker identification. Therefore, it is not necessary to individually handle the input of the speaker information, and the speaker identification information can be obtained without imposing a burden on the speaker. Further, a speaker identification result can be obtained based on two pieces of speaker information of the input biometric information and the input voiceprint information. Therefore, an accurate speaker identification result can be obtained. Further, as described above, speaker identification is performed based on the biometric information and the voiceprint information of the speaker who is uttering now. Therefore, the correspondence between the input voice data and the speaker can be ensured, and highly reliable speaker identification is performed.
[0018]
In the speaker identification device according to one embodiment, the collation result integrating unit compares the S / N ratio of the input biometric information and the S / N ratio of the input voice information with an average value stored in a memory unit in advance. The speaker identification result based on the input information having the higher S / N ratio than the average value is output as the one speaker identification result. Therefore, of the speaker identification result based on the input biometric information and the speaker identification result based on the input voiceprint information, the more likely speaker identification result is output as the final speaker identification result.
[0019]
Further, in the speaker identification device according to one embodiment, the verification result integrating means includes information on the likelihood that the biometric information verification means has obtained the speaker identification result and the voiceprint verification means obtains the speaker identification result. The information of the certainty at the time of comparison is compared with the information of the certainty, and the speaker identification result with the higher certainty is output as the one speaker identification result. Therefore, of the speaker identification result based on the input biometric information and the speaker identification result based on the input voiceprint information, the more likely speaker identification result is output as the final speaker identification result.
[0020]
Further, according to the voice input method of the present invention, when the voice input switch is turned on by the speaker, the biometric information of the speaker is input in response to the on state of the voice input switch, while the voice is input. The audio data generated based on this is input to the subsequent stage. Then, the input biometric information and the registered biometric information registered in the biometric information registration unit are compared and collated, a speaker related to the input voice is identified, and a speaker identification result is output.
[0021]
Thus, the biometric information of the speaker speaking by operating the voice input switch is automatically acquired, and the speaker identification is performed. Therefore, it is not necessary to handle the input of the speaker information individually, and the voice data and the speaker identification information separated by the utterance section can be obtained without putting a burden on the speaker. Further, if the speaker identification information is used for selecting a speaker adaptive dictionary used for speech recognition, it becomes possible to automatically select the speaker adaptive dictionary.
[0022]
According to the speaker identification method of the present invention, the voice input method is executed to obtain a speaker identification result related to the input voice based on the input biometric information, and voiceprint information is extracted from the generated voice data. Is entered. Further, the input voiceprint information and the registered voiceprint information registered in the voiceprint information registration unit are compared and collated, a speaker related to the input voice is identified, and a speaker identification result is output. Then, the speaker identification result based on the input biometric information and the speaker identification result based on the input voiceprint information are integrated, and one speaker identification result is output.
[0023]
Thus, the voice input switch is operated to automatically input the biometric information and the voiceprint information of the speaking speaker, thereby performing the speaker identification. Therefore, it is not necessary to individually handle the input of the speaker information, and the speaker identification information can be obtained without imposing a burden on the speaker. Furthermore, the correspondence between the input voice data and the speaker can be ensured, and highly reliable speaker identification is performed. Further, a speaker identification result can be obtained with high accuracy based on two pieces of speaker information of input biometric information and input voiceprint information.
[0024]
Also, the voice input program of the present invention, when the voice input switch of the voice input device is operated by a speaker, causes the computer to control the voice input means, the biological information input means, the biological information registration means, It is made to function as biometric information matching means.
[0025]
Therefore, it is not necessary to handle the input of the speaker information individually, and the voice data and the speaker identification information separated by the utterance section can be obtained without putting a burden on the speaker. Further, if the speaker identification information is used for selecting a speaker adaptive dictionary used for speech recognition, it becomes possible to automatically select the speaker adaptive dictionary.
[0026]
Also, the speaker identification program of the present invention, when a voice input switch of the speaker identification device is operated by a speaker, causes a computer to input the voice input means, the biological information input means, the biological information of the speaker identification device. It functions as registration means, biometric information matching means, voiceprint information extracting means, voiceprint information registering means, voiceprint matching means, and matching result integrating means.
[0027]
Therefore, it is not necessary to individually handle the input of the speaker information, and the speaker identification information can be obtained without imposing a burden on the speaker. Further, a speaker identification result can be obtained with high accuracy based on two pieces of speaker information of input biometric information and input voiceprint information.
[0028]
Further, the program recording medium of the present invention stores the above-mentioned voice input program. Therefore, by reading and executing the voice input program by the computer, the voice data and the speaker identification information separated by the utterance section can be obtained without imposing a burden on the speaker. Further, if the speaker identification information is used for selecting a speaker adaptive dictionary used for speech recognition, it becomes possible to automatically select the speaker adaptive dictionary.
[0029]
Further, the program recording medium of the present invention stores the above speaker identification program. Therefore, by reading and executing this speaker identification program by a computer, speaker identification information can be obtained without imposing a burden on the speaker. Further, a speaker identification result can be obtained with high accuracy based on two pieces of speaker information of input biometric information and input voiceprint information.
[0030]
BEST MODE FOR CARRYING OUT THE INVENTION
Hereinafter, the present invention will be described in detail with reference to the illustrated embodiments.
[0031]
<First embodiment>
FIG. 1 is a functional block diagram of the voice input device according to the present embodiment. In FIG. 1, reference numeral 1 denotes a biometric information acquisition sensor as the biometric information acquisition means for acquiring biometric information of a speaker, such as a fingerprint reading sensor or a palm print reading sensor. The fingerprint reading sensor or palm print reading sensor or the like does not limit the reading method. For example, an optical device, a device using a pressure-sensitive sheet, a device for measuring a change in capacitance of an electrode, and a field effect transistor An apparatus using the above may be used.
[0032]
Reference numeral 2 denotes a biological information input unit for inputting biological information acquired by the biological information acquisition sensor 1. Reference numeral 3 denotes a biometric information registration unit in which biometric information of one or more persons is registered for each individual. The biological information registered in the biological information registration unit 3 is information corresponding to the biological information acquired by the biological information acquisition sensor 1, for example, a fingerprint or a palm print. Reference numeral 4 denotes a biometric information matching unit that compares input biometric information input from the biometric information input unit 2 with registered biometric information registered in the biometric information registration unit 3 and outputs a speaker identification result.
[0033]
Reference numeral 5 denotes a voice input switch that is turned on when voice is input. Reference numeral 6 denotes a case where when the audio input switch 5 is on, the audio is converted into audio data by a microphone (not shown) or the like, and an A / D (analog / digital) conversion process is performed to generate digital audio data. This is a voice input unit to be input at a later stage. Note that the A / D conversion process in the voice input unit 6 is not an essential process, and analog voice data obtained via a microphone may be directly input.
[0034]
Here, in the functional configuration as described above, the biometric information input unit 2, the biometric information matching unit 4, and the voice input unit 6 are configured by a dedicated LSI (large-scale integrated circuit) element or the like. The biometric information registration unit 3 is configured by a semiconductor memory, a magnetic memory, a storage device, or the like. It should be noted that the present embodiment is not affected whether the elements or the like constituting each of the above sections are one or a plurality of composite elements. Further, each of the above-mentioned units may be replaced by a CPU (Central Processing Unit) or its peripheral devices.
[0035]
Here, in FIG. 1, the biological information acquisition sensor 1 and the voice input switch 5 are illustrated as one module, but this is one of the features of the present embodiment. That is, when the voice input switch 5 is on, the biometric information is acquired by the biometric information acquisition sensor 1 and voice data is generated via the microphone constituting the voice input unit 6. It is preferable that the voice input switch 5 be switched between the on state and the off state while the speaker is pressing down the voice input switch 5 with his / her fingertip and off when the speaker is released. . Further, the utterance end determination means (not shown) is provided, and the utterer presses the voice input switch 5 once immediately before the utterance, and then uses the S / N ratio calculated from the power of the input voice and the like. Then, the end of the utterance section is determined by the utterance end determination means, and the voice input switch 5 may be automatically turned off when the utterance section ends or immediately thereafter.
[0036]
FIG. 2 is a flowchart of a voice input processing operation performed by the voice input device shown in FIG. Hereinafter, the voice input processing operation will be described in detail with reference to FIG. When the voice input switch 5 is switched from the off state to the on state by the speaker, the voice input processing operation starts.
[0037]
In step S1, the biometric information input unit 2 inputs biometric information such as a fingerprint or a palm print acquired by the biometric information acquisition sensor 1 to the biometric information matching unit 4 (biometric information input processing). In step S2, the audio input unit 6 converts the audio into audio data by the microphone or the like, performs A / D conversion processing, generates digital audio data, and inputs the digital audio data to the subsequent stage (audio input processing). In this case, as described above, the A / D conversion processing is not an essential processing, and the converted analog audio data may be directly input via the microphone.
[0038]
In step S3, the biological information collating unit 4 determines whether the voice input switch 5 is on or off. As a result, if it is off, the process proceeds to step S4, while if it is on, the process returns to step S1 to continue the biometric information input process and the voice input process. In step S4, the input biometric information input in step S1 is compared with the registered biometric information registered in the biometric information registration unit 3 by the biometric information matching unit 4. Then, a speaker identification result is output based on the matching result (biological information matching processing). After that, the voice input processing operation ends.
[0039]
The speaker identification result thus output is used, for example, when selecting a speaker adaptive dictionary used for recognizing the input voice data, so that the input voice can be recognized with high accuracy. is there.
[0040]
As described above, in the present embodiment, the voice input switch 5 that is switched from the off state to the on state by the speaker is provided, and when the voice input switch 5 is in the on state, the biological information is obtained by the biological information acquisition sensor 1. The voice is acquired and the voice is input from the voice input unit 6. Then, the biometric information acquired by the biometric information acquisition sensor 1 and the biometric information registered in the biometric information registration unit 3 are collated by the biometric information collation unit 4, and a speaker identification result is obtained based on the collation result. I have.
[0041]
As described above, according to the present embodiment, it is possible to automatically acquire the biological information of the speaking speaker by operating the voice input switch 5. Therefore, it is not necessary to input speaker information individually, and it is possible to obtain the voice data and the speaker identification information separated by the utterance section without putting a burden on the speaker. In addition, since voice data is not used for authentication, the authentication is not affected by ambient noise and the like, and the speaker can be accurately authenticated. In addition, if the speaker identification information is used to select a speaker adaptive dictionary used for speech recognition, the user can automatically select an index without selecting an index as in the case of Via Voice of IBM Japan, Ltd. It becomes possible to select a speaker adaptive dictionary. Further, the voice input switch 5 is operated to acquire the biometric information of the speaking speaker, so that the correspondence between the speech data and the speaking speaker can be obtained, and the security effect can be enhanced.
[0042]
In the first embodiment, the case where the acquisition of the biological information and the input of the voice are continued while the voice input switch 5 is on is described. However, the present embodiment is not limited to this. For example, if it is determined in step S3 of the voice input processing operation shown in FIG. Returning to S2, the input of biometric information may be performed only once while the voice input switch 5 is kept on. In addition, although the voice is input in step S2 after the biometric information is input in step S1, the biometric information input process and the voice input process are preferably performed in parallel. Although the biometric information matching process is performed after the biometric information is input in step S1 and the voice is input in step S2, it is preferable that the biometric information matching process and the voice input process be performed in parallel.
[0043]
<Second embodiment>
This embodiment relates to a speaker identification device using the voice input device according to the first embodiment.
[0044]
FIG. 3 is a functional block diagram of the speaker identification device according to the present embodiment. In FIG. 3, the biometric information acquisition sensor 11, the biometric information input unit 12, the biometric information registration unit 13, the biometric information matching unit 14, the voice input switch 15, and the voice input unit 16 are the biometric information acquisition sensors according to the first embodiment. 1, a biometric information input unit 2, a biometric information registration unit 3, a biometric information collation unit 4, a voice input switch 5, and a voice input unit 6, which constitute a voice input device 17. The detailed description of the voice input device 17 is omitted.
[0045]
However, the voice input unit 16 includes voiceprint information extraction means for analyzing voice data input via the microphone or the like to obtain voiceprint information in addition to the function of the voice input unit 6 in FIG. The input voiceprint information is input to a voiceprint collation unit 18 described later. Note that as the voiceprint information, for example, a feature amount based on a frequency analysis result obtained by FFT (Fast Fourier Transform) or the like may be used.
[0046]
Reference numeral 19 denotes a voiceprint registration unit for registering voiceprint information of one or more persons for each individual. The voiceprint matching unit 18 compares and compares the input voiceprint information input from the voice input unit 16 with the registered voiceprint information registered in the voiceprint registration unit 19, and outputs a speaker identification result. Reference numeral 20 denotes a matching result integrating unit that integrates the speaker identification result from the biometric information matching unit 14 and the speaker identification result from the voiceprint matching unit 18 and outputs one speaker identification result.
[0047]
In the matching result integrating unit 20, for example, the average of the S / N ratios calculated in advance for each of the various biological information input from the biological information input unit 12 and the various audio information input from the audio input unit 16 The value is stored in a semiconductor memory (not shown) or the like. Then, in the verification result integration processing, the S / N ratio of the input biometric information and the S / N ratio of the input voice information are compared with the stored average value, and the S / N ratio larger than the average value is compared. The input information of the one exhibiting the N ratio is adopted, and the speaker identification result based on the adopted input information is output as the one speaker identification result.
[0048]
Alternatively, for example, when the biometric information matching unit 14 and the voiceprint matching unit 18 generate a speaker identification result for each of the biometric information input from the biometric information input unit 12 and the voice information input from the voice input unit 16. , The speaker identification result based on the input information with the higher certainty may be output as the one speaker identification result.
[0049]
Here, in the functional configuration as described above, the biometric information input unit 12, the biometric information matching unit 14, the voice input unit 16, the voiceprint matching unit 18, and the matching result integrating unit 20 are dedicated LSI (large-scale integrated circuit) elements. And so on. The biometric information registration unit 13 and the voiceprint registration unit 19 are configured by a semiconductor memory, a magnetic memory, a storage device, or the like. It should be noted that the present embodiment is not affected whether the elements or the like constituting each of the above sections are one or a plurality of composite elements. Further, each of the above-mentioned units may be replaced by a CPU (Central Processing Unit) or its peripheral devices.
[0050]
FIG. 4 is a flowchart of a speaker identification processing operation executed by the speaker identification device shown in FIG. Hereinafter, the speaker identification processing operation will be described in detail with reference to FIG. When the voice input switch 15 is switched from the off state to the on state by the speaker, the speaker identification processing operation starts.
[0051]
In steps S11 to S14, the voice input device 17 performs biological information input processing and voice input processing in the same manner as in steps S1 to S4 in the voice input processing operation shown in FIG. When is turned off, the biometric information matching process is performed. However, in the voice input process in step S12, when voice data is obtained, the voice data is analyzed and the voiceprint information (input voiceprint information) is extracted.
[0052]
In step S15, the voiceprint verification unit 18 compares and compares the input voiceprint information extracted in step S12 with the registered voiceprint information registered in the voiceprint registration unit 19. Then, a speaker identification result is output based on the collation result (voiceprint information collation processing). In step S16, the speaker identification result obtained as a result of the biometric information matching process in step S14 and the speaker identification result obtained as a result of the voiceprint information matching process in step S15 are output by the matching result integration unit 20. As described above, one speaker identification result is output after being integrated (collation result integration processing). After that, the speaker identification processing operation ends.
[0053]
As described above, in the present embodiment, the voice input device 17 having the same configuration as that of the first embodiment is provided, and when the voice input switch 15 is on, The speaker identification result is obtained based on the collation result between the input biometric information and the registered biometric information. Further, the voice input unit 16 of the voice input device 17 extracts voiceprint information from the obtained voice data, and performs collation between the input voiceprint information and the registered voiceprint information registered in the voiceprint registration unit 19. The voiceprint collating unit 18 obtains a speaker identification result based on the collation result. Then, the matching result integrating unit 20 integrates the speaker identification result based on the input biometric information from the biometric information matching unit 14 and the speaker identification result based on the input voiceprint information from the voiceprint matching unit 18 to form one speech. It outputs the person identification result.
[0054]
As described above, according to the present embodiment, the voice input switch 15 can be operated to automatically acquire the biological information and the voiceprint information of the speaking speaker. Therefore, it is not necessary to individually input speaker information, and speaker identification information can be obtained without imposing a burden on the speaker. Further, a speaker identification result is obtained based on a plurality of pieces of speaker information of the input biometric information and the input voiceprint information. Therefore, a speaker identification result can be obtained with high accuracy. Furthermore, since the voice input switch 15 is operated to acquire the biometric information and voiceprint information of the uttering speaker and perform the speaker identification, the correspondence between the input voice data and the utterer can be ensured, Reliable speaker identification can be performed.
[0055]
In the second embodiment, after performing the biometric information matching process in step S14, the voiceprint information matching process is performed in step S15. It is preferable to carry out in parallel. Further, the biometric information matching process and the voiceprint information matching process are performed separately and independently of the matching result integration process. However, the biometric information matching process and the voiceprint information matching process are performed in the matching result integration process. You can do it.
[0056]
By the way, the biometric information input units 2 and 12, the biometric information registration units 3 and 13, the biometric information matching units 4 and 14, the voice input units 6 and 16, the voiceprint matching unit 18, the voiceprint registration unit 19, and the matching in the above embodiments. The function as the result integrating unit 20 can also be realized by a voice input program and a speaker identification program recorded on a program recording medium. The program recording medium in each of the above embodiments is a program medium composed of a ROM (Read Only Memory). Alternatively, it may be a program medium that is mounted on and read from an external auxiliary storage device. In any case, the program reading means for reading the voice input program or the speaker identification program from the program medium may have a configuration of directly accessing and reading the program medium, or a RAM (random access program). A configuration may be adopted in which the program is downloaded to a program storage area (not shown) provided in the memory, and the program storage area is accessed and read. It is assumed that a download program for downloading from the program medium to the program storage area of the RAM is stored in the main device in advance.
[0057]
Here, the above-mentioned program medium is configured to be separable from the main body side, such as a tape system such as a magnetic tape or a cassette tape, a magnetic disk such as a floppy disk or a hard disk, a CD (compact disk) -ROM, an MO (magneto-optical). Disk system of optical disks such as disk, MD (mini disk), DVD (digital versatile disk), card system such as IC (integrated circuit) card and optical card, mask ROM, EPROM (ultraviolet erasing ROM), EEPROM (electric This is a medium that fixedly carries a program, including a semiconductor memory system such as a temporary erasing ROM) and a flash ROM.
[0058]
In the case where the voice input device and the speaker identification device according to the above-described embodiment have a configuration that can be connected to a communication network including the Internet via a communication I / F, the program medium includes the communication network. It may be a medium that carries the program in a fluid manner by downloading it from, for example. In this case, it is assumed that a download program for downloading from the communication network is stored in the main device in advance. Alternatively, it shall be installed from another recording medium.
[0059]
It should be noted that what is recorded on the recording medium is not limited to only a program, and data can also be recorded.
[0060]
【The invention's effect】
As is clear from the above, in the present invention, when the speaker switches the voice input switch to the side that enables voice input and biometric information acquisition and utters, the voice is converted into voice data and input to the subsequent stage, The biometric information is acquired and input, and the input biometric information is compared with the registered biometric information to output a speaker identification result relating to the input voice, so that the biometric information of the speaking speaker is automatically acquired. Speaker identification can be performed.
[0061]
Therefore, it is not necessary to input speaker information individually, and it is possible to obtain the voice data and the speaker identification information separated by the utterance section without putting a burden on the speaker. In addition, since voice data is not used for authentication, it is possible to accurately perform speaker authentication without being affected by ambient noise. Further, if the speaker identification information is used when selecting a speaker adaptation dictionary used for speech recognition, the speaker adaptation dictionary can be automatically selected.
[0062]
Further, in the present invention, when the speaker switches the voice input switch to the side that enables voice input and biometric information acquisition and utters, the voice data based on the voice and the speaker identification result based on the input biometric information are output. Further, the voiceprint information is extracted from the voice data, the input voiceprint information and the registered voiceprint information are compared and collated, and a speaker identification result is output, and the speaker identification result based on the input biometric information and the input voiceprint information are output. Since one speaker identification result is output by integrating the speaker identification result based on the speaker identification result, the speaker identification can be performed by automatically inputting the biometric information and the voiceprint information of the speaking speaker.
[0063]
Therefore, it is not necessary to individually input speaker information, and speaker identification information can be obtained without imposing a burden on the speaker. Further, a highly accurate speaker identification result can be obtained based on the two pieces of speaker information of the input biometric information and the input voiceprint information. Further, as described above, by performing the speaker identification based on the biometric information and the voiceprint information of the speaker that is currently speaking, the correspondence between the input voice data and the speaker can be ensured. Therefore, highly reliable speaker identification can be performed.
[Brief description of the drawings]
FIG. 1 is a block diagram of a voice input device according to the present invention.
FIG. 2 is a flowchart of a voice input processing operation executed by the voice input device shown in FIG. 1;
FIG. 3 is a block diagram of the speaker identification device according to the present invention.
FIG. 4 is a flowchart of a speaker identification processing operation executed by the speaker identification device shown in FIG. 3;
[Explanation of symbols]
1, 11 ... biological information acquisition sensor,
2, 12 ... biological information input unit,
3, 13 ... biological information registration unit,
4, 14 ... biometric information collation unit,
5, 15 ... voice input switch,
6, 16 ... voice input unit,
17 voice input device,
18. Voiceprint collation unit
19 ... Voiceprint registration section,
20: Matching result integration unit.

Claims (15)

Voice input means for generating voice data based on the voice of the speaker and inputting the voice data to a subsequent stage;
Biological information acquisition means for acquiring the biological information of the speaker,
Biological information input means for inputting biological information acquired by the biological information acquisition means,
Biometric information registration means in which biometric information is registered for each individual;
Biometric information matching means for comparing and comparing input biometric information input from the biometric information input means and registered biometric information registered in the biometric information registration means, and identifying a speaker related to the input voice,
It is operated by a speaker to determine whether or not the voice input by the voice input unit is possible and whether or not the biological information can be obtained by the biological information obtaining unit, and that the biological information can be obtained when the voice input is possible. A voice input device comprising a voice input switch for switching. The voice input device according to claim 1,
A voice input device, wherein the biometric information includes a fingerprint or a palm print. The voice input device according to claim 1,
The voice input switch is in an on state that allows the voice input and the acquisition of biometric information while being pressed, and is in an off state that disables the voice input and the acquisition of biometric information while not pressed. A voice input device, comprising: The voice input device according to claim 1,
Speech end determination means for determining the end of the utterance based on the input voice,
When the voice input switch is pressed, the voice input and the biological information acquisition are turned on to enable the voice input and the biological information acquisition, and when the speech termination determination unit determines that the utterance has ended, the voice input and the biological information acquisition are performed. A voice input device, which is in an off state for disabling. The voice input device according to claim 3 or 4,
The voice input device, wherein the biometric information obtaining means continuously obtains biometric information while the voice input switch is on. The voice input device according to claim 3 or 4,
The voice input device, wherein the biometric information obtaining means obtains biometric information only once when the voice input switch is turned on. A voice input device according to any one of claims 1 to 6,
Voiceprint information extraction means provided in the voice input means of the voice input device, for extracting voiceprint information from the generated voice data;
Voiceprint information registration means in which voiceprint information is registered for each individual;
Voiceprint matching means for comparing and collating input voiceprint information input from the voice input means with registered voiceprint information registered in the voiceprint information registration means, and identifying a speaker related to the input voice;
The apparatus further comprises a verification result integration unit that integrates the speaker identification result by the biometric information verification unit of the voice input device and the speaker identification result by the voiceprint verification unit, and outputs one speaker identification result. Speaker identification device. The speaker identification device according to claim 7,
The collation result integration means includes:
The average value of the S / N ratio of the biometric information and the average of the S / N ratio of the voice information calculated in advance based on various biometric information and voice information input from the biometric information input unit and the voice input unit And the value is stored in the memory means,
At the time of speaker identification, the S / N ratio of the input biometric information and the S / N ratio of the input voice information are compared with the stored average value, and the one that presents a higher S / N ratio than the average value is compared. A speaker identification device configured to output a speaker identification result based on input information as the one speaker identification result. The speaker identification device according to claim 7,
The collation result integration means includes:
By comparing the information on the certainty when the biometric information matching means of the voice input device has obtained the speaker identification result with the information on the certainty when the voiceprint matching means has obtained the speaker identification result, A speaker identification device characterized in that a speaker identification result with a higher likelihood is output as the one speaker identification result. A biometric information input step of acquiring and inputting biometric information of the speaker in response to the on state of the voice input switch;
In response to the ON state of the voice input switch, a voice input step of generating voice data based on the voice of the speaker and inputting the generated voice data to a subsequent stage;
A voice input method comprising comparing and comparing the input biometric information input with the registered biometric information registered in the biometric information registration means to identify a speaker associated with the input voice. . A voice input step of executing the voice input method according to claim 10 to obtain a speaker identification result of the input voice based on the input biometric information, and extracting and inputting voiceprint information from the generated voice data. ,
A voiceprint verification step of comparing and collating the input voiceprint information input with the registered voiceprint information registered in the voiceprint information registration means to identify a speaker related to the input voice;
A speaker integrating step of integrating the speaker identification result based on the input biometric information and the speaker identification result based on the input voiceprint information to output one speaker identification result. Identification method. Computer
When the voice input switch according to claim 1 is operated by a speaker, the switch functions as the voice input unit, the biometric information input unit, the biometric information registration unit, and the biometric information matching unit according to claim 1. Voice input program. Computer
When the voice input switch according to claim 7 is operated by a speaker, the voice input means, biometric information input means, biometric information registration means, biometric information matching means, voiceprint information extraction means, voiceprint according to claim 7 A speaker identification program functioning as information registration means, voiceprint collation means and collation result integration means. A computer-readable program recording medium on which the voice input program according to claim 12 is recorded. A computer-readable program recording medium on which the speaker identification program according to claim 13 is recorded.

Images