JP2004101901A - Speech interaction system and speech interaction program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To achieve interaction according to a speaker by accurately recognizing whether a speaker is a registered user or a new user. <P>SOLUTION: A user database storage part 4, which stores a user speech database 11 for speaker recognition and a personal information database 12, is prepared. When speech interaction with the user is carried out, a speaker recognition part 2 identifies the user from an inputted speech signal; and an interaction control part 23 reads out user information corresponding to the identified speaker and selects an interaction scenario corresponding to the user out of a plurality of interaction scenarios according to the read-out user information. Then the interaction control part 23 realizes interaction using the interaction scenario selected corresponding to the user. <P>COPYRIGHT: (C)2004,JPO

Description

[0001]
TECHNICAL FIELD OF THE INVENTION
The present invention relates to a voice dialogue apparatus and a voice dialogue program that recognize a user's speaker when realizing a voice dialogue with the user and change the content of the dialogue according to the result of the speaker recognition.
[0002]
[Prior art]
2. Description of the Related Art Conventionally, there has been known a voice interaction system that emits a voice to a user and further inputs a voice emitted from the user to realize a dialogue with the user. In this conventional voice dialogue system, it is often the case that the current user is not recognized, and the dialogue is executed regardless of whether the user is a new user or a user who has used in the past.
[0003]
On the other hand, in another conventional voice interaction system, for example, the user is recognized by prompting the user to input a key such as a password.
[0004]
[Patent Document 1]
JP-A-63-85698
[0005]
[Problems to be solved by the invention]
However, many of the above-described conventional voice dialogue systems perform a dialogue without recognizing who the current talker is, and therefore do not refer to the unique information or knowledge of the talker. Therefore, in the conventional voice dialogue system, there is a problem in terms of performance and efficiency in that effective dialogue with the user does not progress, and the content of the previous dialogue is repeatedly executed.
[0006]
Further, in another conventional voice interaction system, the user is identified by prompting the user to enter a password. However, there is a problem that an operation burden is imposed on the user when starting the dialogue.
[0007]
Therefore, the present invention has been proposed in view of the above-described situation, and makes it possible to accurately perform speaker recognition as to whether the user is a registered user or a new user, and perform a dialog according to the speaker. It is an object of the present invention to provide a spoken dialogue apparatus and a spoken dialogue program to be realized.
[0008]
[Means for Solving the Problems]
According to the present invention, a database storage unit for storing user information about a user is prepared, and when performing a voice dialogue with the user, a speaker is identified from an input voice signal, and a user corresponding to the identified speaker is identified. The information is read, and an interaction scenario corresponding to the user is selected from the plurality of interaction scenarios based on the read user information. The present invention solves the above-described problem by realizing a dialog using a dialog scenario selected according to a user.
[0009]
BEST MODE FOR CARRYING OUT THE INVENTION
Hereinafter, embodiments of the present invention will be described with reference to the drawings.
[0010]
[Configuration of voice interaction device]
The present invention is applied to, for example, a spoken dialogue apparatus 1 configured as shown in FIG. The speech dialogue apparatus 1 stores a speech dialogue program in a storage mechanism (not shown), and executes the speech dialogue program by a computer including a CPU (Central Processing Unit) (not shown) and the like. The function module of each section of the section 3 and the voice storage section 5 is configured. The contents of the user database storage unit 4 and the dialog scenario storage unit 21 in the voice dialogue unit 3 are read by each functional module configured by executing the voice dialogue program, and information update and new information registration are performed. Processing is performed.
[0011]
The voice interactive device 1 is provided with a sound collecting mechanism (not shown) such as a microphone toward the front position of the device, and a voice emitted from the user via the sound collecting mechanism is recognized as a voice signal of an electric signal by a speaker. It is input to the section 2 and the voice dialogue section 3.
[0012]
In the spoken dialogue apparatus 1, a speaker recognition user speech database 11 and a personal information database 12 are prepared in advance and stored in the user database storage unit 4 in order to manage information about the user. Further, in the voice interaction device 1, a plurality of interaction scenarios for realizing different interactions depending on the user are prepared in advance, and the interaction scenarios are stored in the interaction scenario storage unit 21. In this example, a dialog scenario 31 for a new user and a dialog scenario 32 for a registered user are prepared as the dialog scenario.
[0013]
The user database storage unit 4 is formed of a recording medium such as a hard disk capable of storing a large amount of data. In the speaker recognition user voice database 11 stored in the user database storage unit 4, voices of users who have used the voice interactive device 1 in the past are stored as voice signals. That is, the voice signal stored in the speaker recognition user voice database 11 is registered and stored in association with the name of the personal information in the personal information database 12 and the like.
[0014]
In this example, the voice signal stored in the speaker recognition user voice database 11 includes the name of the user. The speaker recognition user voice database 11 may store a plurality of voice signals having the same name. However, since the frequency characteristics of voice signals differ from individual to individual depending on the pitch of the voice, a speaker described later is used. The recognizing unit 2 can recognize it.
[0015]
The personal information database 12 stores various types of personal information including names of users who have used the voice interaction device 1 in the past. The various types of personal information are the usage history of the voice interaction device 1 and the purpose of use, and vary depending on the use of the voice interaction device 1. A specific example of the personal information will be described in a specific operation of the voice interaction device 1 described later.
[0016]
The speaker recognition unit 2 uses the contents of the speaker recognition user voice database 11 stored in the user database storage unit 4, a voice signal input from the outside, and a voice recognition result from the voice dialogue unit 3 to perform speaker recognition. do.
[0017]
At this time, when the speech recognition result from the speech dialogue unit 3 is input in, for example, a text format, the speaker recognition unit 2 extracts a speech signal of a text similar to the text from the speaker recognition user speech database 11. Performs user identification processing. Next, the speaker recognizing unit 2 performs a speaker matching process of determining the single speaker from the user candidates by comparing the extracted voice signal as the user candidate with the input voice signal.
[0018]
As a result, if the speaker is determined as a result of the speaker recognition processing, the speaker recognizing unit 2 sends information to that effect to the voice dialogue unit 3 and, if it is determined that the speaker is a new user, notifies the voice Send to dialogue unit 3.
[0019]
Here, for example, even if N people are extracted as speaker candidates, when other people than the N people actually utter, the most similar voice of the N people is obtained only by the speaker identification processing. Although a person having possession is selected as a candidate, in order to determine that the person is another person (or a new person) other than N, speaker verification processing is performed after speaker identification processing for extracting speaker candidates, and N , Or a person other than N (or a new person).
[0020]
The voice interaction device 1 further includes a voice storage unit 5 that receives a voice recognition result of the voice signal input by the voice interaction unit 3 and registers the voice signal in the speaker recognition user voice database 11. When a voice signal from the voice storage unit 5 is input, the speaker recognition user voice database 11 is associated with the personal information stored in the personal information database 12 and can be referred to when the speaker recognition unit 2 performs the speaker recognition process. And
[0021]
The voice dialogue unit 3 includes a voice recognition unit 22, a dialogue control unit 23, and a voice synthesis unit 24 that externally output a user's voice signal. In the voice interaction unit 3, when a voice signal is input by the voice recognition unit 22, voice recognition is performed by the voice recognition unit 22.
[0022]
At this time, the voice recognition unit 22 performs voice recognition by comparing the input voice signal with a prepared voice identification database, and outputs the voice recognition result to the speaker recognition unit 2, the voice storage unit 5, or the dialog control unit. Send to 23. Here, the start timing of voice recognition by the voice recognition unit 22 is a timing at which the sound pressure level becomes equal to or higher than a predetermined value and the sound pressure level continues for a predetermined time or longer. Then, the speech recognition unit 22 performs a Fourier transform or the like on the speech signal input after the start timing of the speech recognition to extract a speech feature amount, and performs speech recognition using the speech feature amount.
[0023]
When causing the speaker recognition unit 2 to perform speaker recognition, the speech recognition unit 22 sends the speech recognition result to the speaker recognition unit 2 in a text format. At this time, the voice recognition unit 22 sends a plurality of texts having higher scores (certainty) of voice recognition to the speaker recognition unit 2 as voice recognition results. When the speaker is determined by the speaker recognizing unit 2 as a result of the speaker recognition, the voice recognizing unit 22 sends the voice signal at that time to the voice storage unit 5. As a result, the voice storage unit 5 causes the speaker recognition user voice database 11 to newly register a voice signal or update a voice signal already registered.
[0024]
The voice synthesizing unit 24 is connected to a sound emitting mechanism such as a speaker, and emits voices of various contents toward the user under the control of the dialog control unit 23. In this example, the text generated by the dialogue control unit 23 with reference to the dialogue scenario storage unit 21 and the personal information database 12 is sent to the speech synthesis unit 24, and the speech synthesis unit 24 synthesizes the speech to be issued to the user. Voice will be generated.
[0025]
The dialogue control unit 23 selects a dialogue scenario corresponding to the speaker from the dialogue scenario storage unit 21 according to the speaker recognition result from the speaker recognition unit 2. At this time, the interaction control unit 23 refers to the personal information database 12 and reads out an interaction scenario for each user. Then, the conversation control unit 23 refers to the personal information database 12 and controls the speech synthesis unit 24 according to the speech recognition result from the speech recognition unit 22 to emit a sound corresponding to the speaker.
[0026]
When a speaker recognition result indicating that the user is a new user is input from the speaker recognition unit 2, the dialog control unit 23 reads the dialog scenario 31 for a new user from the dialog scenario storage unit 21 and emits sound. When the speaker control unit 23 receives a speaker recognition result indicating that the user is a registered user in the user database storage unit 4 from the speaker recognition unit 2, the dialog scenario for registered users is input from the dialog scenario storage unit 21. 32 is read and sound is emitted. Here, the registered user interaction scenario 32 may be prepared for each user registered in the user database storage unit 4, or may be prepared by setting a predetermined category or the like.
[0027]
In such a voice interaction device 1, when the speaker recognition unit 2 recognizes a registered user, the voice interaction proceeds with the registered user interaction scenario 32 while referring to the personal information database 12. Thus, the voice interaction device 1 can function as a voice interaction agent corresponding to an individual.
[0028]
That is, in the voice dialogue apparatus 1, in response to input of voice from the user, the voice recognition unit 22 performs voice recognition to convert a voice signal into text, and causes the text to be input to the dialogue control unit 23. Thus, the dialogue control unit 23 refers to the personal information database 12 and the dialogue scenario to generate a text to be answered from the voice dialogue apparatus 1 to the user. As a result, the voice synthesizer 24 inputs the text to be replied, converts the text to voice, and replies to the user.
[0029]
Further, in the voice dialogue apparatus 1, when the speaker recognition unit 2 recognizes a new user, the voice dialogue proceeds using the dialogue scenario 31 for a new user. As a result, in the voice interaction device 1, the registered user and the new user are distinguished from each other, and the voice interaction proceeds. Here, in the case of a new user, information on the user is newly registered in the speaker recognition user voice database 11 and the personal information database 12 during or after the voice conversation.
[0030]
[Other Configurations of Voice Dialogue Device]
Next, another voice interactive device 40 to which the present invention is applied will be described with reference to FIG. The same parts as those of the voice interaction device 1 shown in FIG. 1 described above are denoted by the same reference numerals, and detailed description thereof will be omitted.
[0031]
The voice interaction device 40 illustrated in FIG. 2 differs from the voice interaction device 1 illustrated in FIG. 1 in that a face image of a user is captured when recognizing a speaker, and speaker recognition is performed using the face image.
[0032]
This voice interaction device 40 includes a face image recognition database 41, a face image recognition unit 42, and a user recognition unit 43 in addition to the voice interaction device 1 shown in FIG. Such a voice interaction device 40 includes a camera mechanism (not shown) having a viewing angle at the user's standing position, and inputs face image data from the camera mechanism to the face image recognition unit 42. Here, the input timing of the face image includes, for example, a timing at which a user's voice is input to the speaker recognition unit 2 and the speaker recognition unit 2 starts speaker recognition.
[0033]
When face image data is input, the face image recognition unit 42 extracts a face feature amount from the face image data. Then, the face image recognition unit 42 compares the extracted face feature amounts with the face feature amounts of a plurality of users registered in the face image recognition database 41 and performs matching to thereby determine the current voice interaction device 40. Recognize a plurality of user candidates trying to use. Then, the face image recognition unit 42 creates matching scores for a plurality of user candidates and sends the matching scores to the user recognition unit 43 as face image recognition results. In the face image recognition database 41, the facial feature amounts of users who have used the voice interaction device 40 in the past are stored in association with personal information.
[0034]
Further, the speaker recognition result is sent from the speaker recognition unit 2 to the user recognition unit 43. In the voice interaction device 40, unlike the case of the voice interaction device 1, in order to perform user recognition comprehensively with a matching score using a face image by the face image recognition unit 42, a voice signal is output by the speaker recognition unit 2. A matching score is created for each of the extracted user candidates and sent to the user recognition unit 43.
[0035]
The user recognition unit 43 performs user recognition using the matching score from the speaker recognition unit 2 and the matching score from the face image recognition unit 42. At this time, the user recognizing unit 43 performs a composite operation using the matching score from the face image recognizing unit 42 and the matching score from the speaker recognizing unit 2 for each user candidate, and exceeds a predetermined threshold. The user with the matching score obtained is determined as the speaker.
[0036]
In such a voice interaction device 40, since the speaker is recognized using both the face image and the voice signal, the speaker recognition rate is improved as compared with the case where the speaker is recognized using only the voice signal.
[0037]
[Specific operation of voice interaction device]
Next, a specific voice interaction process by the above-described voice interaction device will be described with reference to FIG. In the description with reference to FIG. 3, the voice dialogue processing by the voice dialogue apparatus 40 shown in FIG. 2 will be described.
[0038]
Further, in this example, a description will be given of, for example, the voice interaction device 40 acting as a reception guide of a hospital. That is, the voice interaction device 40 executes a voice interaction using the new user interaction scenario 31 for the first visitor, and a registered user interaction scenario 32 for the visitor who used the voice interaction device 40 in the past. A case in which a voice dialogue is executed using will be described.
[0039]
First, when the voice interaction device 40 recognizes that a visitor is present at the user's standing position, the process proceeds to step S1, and the dialogue control unit 23 prompts the user to ask "What is your name?" The combining unit 24 is controlled, and the process proceeds to step S2.
[0040]
In step S2, in response to the inquiry in step S1, if the visitor claims to be "Nishiyama," the audio signal is converted into digital data via an input device such as a microphone and an A / D converter. The information is acquired by the voice recognition unit 22 and the speaker recognition unit 2, and the process proceeds to step S3.
[0041]
In step S3, the voice feature is extracted by performing voice recognition processing using the voice signal input in step S2 by the voice recognition unit 22, and “Nishiyama” is obtained as text, and the voice is recognized as a user candidate. Sent to the person recognition unit 2 and proceeds to step S4. In this example, the name of the speaker is used as the authentication ID. In addition, as a voice recognition method by the voice recognition unit 22, HMM (Hidden Markov Model), DP (Dynamic Programming) matching, or another method is used.
[0042]
In step S4, the speaker recognizing unit 2 determines whether or not the user candidate text acquired from the voice recognizing unit 22 in step S3 is registered in the speaker recognizing user voice database 11. At this time, in addition to the text “Nishiyama”, the speaker recognition unit 2 searches the speaker recognition user voice database 11 for other texts similar to “Nishiyama” as user candidates. When the speaker recognition unit 2 determines that the text as the user candidate exists in the user voice database 11 for speaker recognition, the process proceeds to step S5, and when it is determined that the text does not exist, the process proceeds to step S6.
[0043]
In step S5, the speaker recognizing unit 2 reads out and acquires a speech signal associated with a text (name) similar to the text (name) input in step S3 from the speaker recognition user speech database 11. The process proceeds to step S7.
[0044]
In step S7, the speaker recognition unit 2 matches the voice signal obtained in step S2 with the voice signal read out and obtained in step S5 to create a matching score Ms1 and sends it to the user recognition unit 43. To step S8.
[0045]
In step S8, the user recognizing unit 43 uses the matching score Ms1 using the voice signal input in step S7 and the matching score Ms2 using the face image to perform a compound operation on both matching scores. A total score Mss is created, and the process proceeds to step S9.
[0046]
Here, in steps S1 to S3, the case has been described in which the speaker recognition unit 2 acquires a text corresponding to a voice signal. However, in parallel with the processing in steps S1 to S3, a user candidate using a face image is Extract. At this time, in the voice interaction device 40, the face image is input by the face image recognition unit 42 to extract the feature amount of the face image, and the extracted face feature amount and the face feature amount stored in the face image recognition database 41 are compared. And sends a matching score Ms2 for a plurality of user candidates to the user recognition unit 43.
[0047]
In step S9, the user recognition unit 43 determines whether or not the total score Mss calculated in step S8 is larger than a preset threshold. Here, the threshold value is set in advance to a matching score that specifies a registered speaker when a composite operation is performed on the matching score Ms1 using the audio signal and the matching score Ms2 using the face image. This threshold value is calculated by the method of calculating the matching score Ms1 created by the speaker recognizing unit 2 or the matching score Ms2 created by the face image recognizing unit 42 when designing the system of the voice interactive device 40, It changes depending on the calculation method of the total score Mss.
[0048]
If it is determined in step S9 that the total score Mss is not larger than the threshold, the process proceeds to step S10 in order to identify a new user or another registered speaker, and the total score Mss is determined. If it is determined that the value is larger than the value, it is determined that the registered speaker has been specified, and the process proceeds to step S12.
[0049]
In step S10, the user recognizing unit 43 determines whether or not there is another text that is a speaker candidate. If it is determined that there is another text, the text is processed in the previous steps S5 to S8. The process returns to step S5 excluding the text that has been described, and returns to step S4 if it is determined that the text does not exist.
[0050]
Then, in the voice interaction apparatus 40, as long as the text acquired by the speaker recognition unit 2 exists as a user candidate, the processing of step S5, step S7 to step S10 is repeated, and there is no text as a user candidate. When the determination is made, the process proceeds to step S4, and the process proceeds from step S4 to step S6.
[0051]
In step S6, as a result of repeating the processing of step S5 and steps S7 to S10, there is no user candidate text ("Nishiyama" and a text similar thereto) in this example. It is confirmed whether or not it is not a new visitor, and the process returns to step S1. If it is a new visitor, the new speaker ID is registered in the personal information database 12, and the process proceeds to step S13.
[0052]
On the other hand, in step S12, when the user recognition unit 43 determines that there is a text of the user candidate whose overall score Mss exceeds the threshold, in step S12, the text of the user candidate is interacted with by the speaker recognition unit 2. Send to control unit 23. Then, the dialog control unit 23 reads the personal information corresponding to the text of the user candidate with reference to the personal information database 12, and proceeds to step S13.
[0053]
In step S13, the dialogue control unit 23 selects and reads out the dialogue scenario 31 for new users or the dialogue scenario 32 for registered users from the dialogue scenario storage unit 21. At this time, the dialogue control unit 23 reads the dialogue scenario 32 for the registered user when the personal information is obtained in step S12, proceeds to step S14, and when the new speaker ID is obtained in step S11. Reads out the new user interaction scenario 31, and proceeds to step S14.
[0054]
In step S14, the dialogue control unit 23 sends text data to the speech synthesis unit 24 so as to synthesize speech in accordance with the dialogue scenario selected in step S13. In step S15, the speech is released in step S14. A voice corresponding to the sound is input and the voice recognition unit 22 performs voice recognition.
[0055]
By executing Steps S14 and S15, the dialogue control unit 23 uses the dialogue scenario 31 for new users to voice-communicate personal information such as the name, date of birth, address, contact information, and history of the visitor. Get by. On the other hand, when the registered user interaction scenario 32 is used, the conversation is advanced while referring to the previous visit record and the like from the personal information database 12 without inquiring about the same personal information as the new visitor.
[0056]
Then, in step S16 when step S14 and step S15 are completed, the dialog control unit 23 determines whether or not the dialog using the dialog scenario has been completed. If it is determined that the conversation has not ended, the process proceeds to step S17, and the processes of steps S13 to S15 are repeated with reference to the personal information database 12.
[0057]
On the other hand, if it is determined in step S16 that the dialog has ended, the contents of the speaker recognition user voice database 11 and the personal information database 12 are updated to reflect the voice dialog in steps S14 and S15. That is, the dialog control unit 23 stores the latest user voice in the personal information database 12 when it is determined that the dialog has ended, and controls the voice storage unit 5 to update the voice to the latest user voice. Further, the dialog control unit 23 rewrites the contents of the personal information database 12 so as to update the user's personal information according to the contents of the dialog.
[0058]
[Effects of Embodiment]
As described above in detail, according to the voice interaction apparatuses 1 and 40 to which the present invention is applied, whether a registered user or a new user is recognized from the input voice and / or face image, and the registered user And a new user can select a dialog scenario to distinguish the contents of the dialog. Further, according to the voice interaction devices 1 and 40, the registered user can change the content of the interaction according to the personal information with reference to the personal information database 12, and further realize the interaction corresponding to the user. be able to.
[0059]
That is, according to the voice interaction devices 1 and 40, by performing speaker recognition before starting a specific interaction, if the user is a registered user, the same content that was previously interacted does not interact. The dialog scenario can be changed so that the information and knowledge about the user obtained in the previous dialog are reflected in the next dialog, and an intelligent function closer to a human dialog can be provided.
[0060]
Further, according to the voice interaction devices 1 and 40, information for identifying the interaction scenario is registered in the personal information database 12 in association with the personal information, and an interaction scenario for each individual is prepared in the interaction scenario storage unit 21. You can keep it. By doing so, the dialog control unit 23 can select an appropriate dialog scenario from the dialog scenario storage unit 21 from personal information corresponding to the speaker, according to the speaker recognition result.
[0061]
Thereby, for example, information such as the number of visits and the purpose of the visit is acquired as personal information, and the dialog control unit 23 confirms in step S14 and step S15 whether the visitor has visited the hospital this time. In the next step S13, it is possible to start a dialogue using a visit dialogue scenario. In addition to this effect, the voice dialogue apparatuses 1 and 40 prepare a dialogue scenario for a new user, so that a natural dialogue can be realized even for a new user.
[0062]
Further, according to the voice interaction devices 1 and 40, it is not necessary to prompt for a password or the like when recognizing a speaker, and the burden on the user can be reduced.
[0063]
Further, according to the voice interaction devices 1 and 40, a voice signal for matching in the speaker recognition unit 2 is obtained from the speaker recognition user voice database 11 from the text obtained by performing voice recognition by the voice recognition unit 22. Therefore, it is not necessary to perform matching between the input voice signal and all the voice signals stored in the speaker recognition user voice database 11, and the amount of calculation in the speaker recognition unit 2 can be reduced.
[0064]
Furthermore, according to the voice interaction devices 1 and 40, the start of the utterance of the user may be determined only from the level of the voice signal. Further, the face image recognition unit 42 detects the movement of the mouth of the user's face image. Detection may be performed to determine the start of the utterance of the user. This makes it possible to accurately determine the start and end of the utterance of the user and start and end the speech recognition and the speaker recognition in a timely manner. This can also improve the accuracy of voice recognition and speaker recognition, and reduce erroneous recognition.
[0065]
Note that the above embodiment is an example of the present invention. For this reason, the present invention is not limited to the above-described embodiment, and other than the present embodiment, various modifications may be made according to the design and the like within a range not departing from the technical idea according to the present invention. Can be changed.
[0066]
That is, in the above-described example, the case where the voice dialogue devices 1 and 40 are applied to reception at a hospital has been described as a specific example of use. However, the present invention is not limited to this. The same effects as described above can of course be achieved even when applied to an application to an agent, an application to an electronic approval system using speaker verification on a mobile phone, and the like.
[0067]
【The invention's effect】
According to the present invention, speaker recognition is performed from voice, a dialogue scenario is selected according to user information, and a voice dialogue is performed according to the dialogue scenario. It is possible to accurately perform speaker recognition as to whether the user is a new user or a new user, and realize a conversation corresponding to the speaker.
[Brief description of the drawings]
FIG. 1 is a block diagram illustrating a configuration of a voice interaction device to which the present invention has been applied.
FIG. 2 is a block diagram showing a configuration of another voice interactive device to which the present invention is applied.
FIG. 3 is a flowchart illustrating a specific process of the voice interaction device to which the present invention is applied;
[Explanation of symbols]
1,40 spoken dialogue device
2 Speaker recognition unit
3 Voice Dialogue Department
4 User database storage
5 Voice storage unit
11 User speech database for speaker recognition
12 Personal information database
21 Dialogue scenario storage unit
22 Voice Recognition Unit
23 Dialogue control unit
24 Voice synthesis unit
31 Dialogue scenario for new users
32 Dialogue scenarios for registered users
41 Face Image Recognition Database
42 face image recognition unit
43 User Recognition Unit

Claims (10)

Voice input means for receiving a voice from the user and generating a voice signal;
Audio output means for emitting audio to the user,
Speaker recognition means for performing speaker identification from the voice signal generated by the voice input means, database storage means for storing a user database storing user information about the user,
Information corresponding to the speaker identified by the speaker recognition means is read from the database storage means, and based on the read user information, an interaction scenario corresponding to the user is selected from a plurality of interaction scenarios. A dialogue control unit that controls the voice output unit so as to output a voice according to a dialogue scenario. The database storage means includes, as the user database, a user voice database in which voice signals previously input by the voice input means are stored and registered for each user, and a personal information database in which personal information is registered for each user. Is remembered,
The speaker recognition unit performs speaker identification by comparing a voice signal registered in the user voice database with a voice signal generated by the voice input unit, and the dialog control unit outputs a speaker identification result. The voice interaction device according to claim 1, wherein a dialog scenario corresponding to a user is selected by referring to the personal information database based on the personal information database. The speaker recognizing means performs voice recognition using the voice signal generated by the voice input means, and extracts a user candidate voice signal registered in the user voice database from the voice recognition result. 3. The voice interaction apparatus according to claim 2, wherein the voice recognition unit compares the voice signal generated by the voice input unit with the voice signal generated by the voice input unit to perform speaker recognition. As a result of performing the speaker recognition, the speaker recognition unit determines whether the user is a registered user or a new user,
When it is determined that the user is a registered user by the speaker recognition means, the dialog control means selects a scenario for the registered user as a dialog scenario and performs a voice dialogue. 2. The voice interaction apparatus according to claim 1, wherein when it is determined that the user is a user, a scenario for a new user is selected as the interaction scenario to have a voice interaction. The speaker recognition unit further includes a face image capturing unit that captures a face image of the user, and a voice signal input by the voice input unit in the past is stored in a user voice database stored and registered for each user. The voice signal is collated with the voice signal generated by the voice input means to perform speaker identification, and a face image previously registered in a face image recognition database and a face image captured by the face image capturing means are compared. The voice dialogue according to claim 1, wherein the speaker is identified by collating, and the user is recognized based on the speaker identification result using the voice signal and the speaker identification result using the face image. apparatus. When conducting a voice dialogue with the user,
Identify the speaker from the input voice signal,
Reading information corresponding to the identified speaker from a database storage unit that stores user information about the user;
Based on the read user information, select an interaction scenario corresponding to the user from among a plurality of interaction scenarios,
A speech dialogue program for causing a computer to execute a process of outputting speech according to the above-mentioned dialogue scenario. The voice signal is read from the user voice database in which the voice signal is registered for each user, the speaker is identified by comparing the voice signal with the input voice signal,
7. The voice according to claim 6, wherein the computer is caused to execute a process of selecting a conversation scenario corresponding to the user by referring to a personal information database in which personal information is registered for each user based on a speaker identification result. Dialogue program. The voice recognition is performed using the input voice signal, and a voice signal to be a user candidate registered in the user voice database is extracted from the voice recognition result,
The computer-readable storage medium according to claim 7, wherein the computer performs a process of performing speaker recognition by comparing the extracted voice signal with the input voice signal. By performing speaker recognition, it is determined whether the user is already registered or a new user,
When it is determined that the user is a registered user, a scenario for the registered user is selected as a dialog scenario, and a voice dialogue is performed.
7. The voice dialogue program according to claim 6, wherein, when it is determined that the user is a new user, a computer performs a process of selecting a scenario for a new user as a dialogue scenario and performing voice dialogue. The voice signal is stored for each user, and the voice signal registered in a user voice database registered and collated with the input voice signal is used to identify a speaker, and a face image registered in a face image recognition database in advance and image pickup is performed. The speaker is identified by comparing it with the
The computer-readable storage medium according to claim 6, wherein the program causes a computer to execute a process of recognizing a user based on a speaker identification result using a voice signal and a speaker identification result using a face image.

Images