JPH10124087A - Voice interactive device and interactive method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To retrieve an object item by changing the order of generations by providing a voice signal accumulating means, voice recognizing through changing the order of inputted voice signals and narrowing the number of voice recognition objects against the word previously generated. SOLUTION: An interactive control section 105 instructs an input voice control section 101 and an input voice accumulating section 102 to accumulate inputted voice signals. Then, the section 101 outputs the inputted voice signal to the section 102 by the instruction and the section 102 starts to accumulate the inputted voice signals. Then, the vocabulary of the recognition objects corresponding to the words generated previously is reduced employing the voice recognition result of the words generated later and voice recognition is conducted. Thus, the flow of interactive operations, in which main information is inputted first and then, auxiliary information is inputted employing a voice recognition section having same performance, is established and an object item is retrieved by changing the order of generation.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a spoken dialogue apparatus and a spoken dialogue method using a speech recognition technique and a speech synthesis technique.

[0002]

2. Description of the Related Art In a device capable of voice communication with a person,
If the number of items included in the group containing a certain target item to be selected exceeds the processing capability of the speech recognition unit, a portion including the target item is input before the target item name is input by voice. It is necessary to input words representing a group, specify a search target in the group, and narrow down the number of words to be subjected to speech recognition.

For example, when a golf course is searched using an item search function for setting a destination using a voice dialog realized in a car navigation apparatus having a voice dialog function, a golf course to be searched is used. If the total number of items is 2,000 in Japan and the maximum processing capacity of the voice recognition unit is 100 words, it is impossible to search all golf course names in Japan as voice recognition targets at once. .

[0004] If the number of facilities in each prefecture is less than 100 when the categories are classified by prefecture, the user is required to input the name of the prefecture before inputting the target golf course name.
By narrowing down the voice recognition targets for each prefecture and then uttering the target facility name, even if the total number of items exceeds the maximum processing capacity of the voice recognition unit, search for the target facility name from all the items Becomes possible.

[0005] Conventionally, as such a speech dialogue device,
For example, there was one as shown in FIGS. FIG.
FIG. 8 is a block diagram showing a configuration of a conventional voice interaction device, and FIG. 8 is a flowchart showing a flow of voice interaction by the voice interaction device shown in FIG.

First, the configuration of a conventional voice interaction apparatus will be described with reference to FIG. In FIG. 7, reference numeral 303 denotes a sound analysis unit for inputting a voice signal and analyzing the input voice signal to obtain a characteristic parameter; A voice recognition unit that performs voice recognition by collating with a recognition dictionary, a dialog control unit 305 that controls voice dialogue, and a dialogue 306 that stores information on the flow of the voice dialogue based on the operation of the user and the result of voice recognition. This is a control information storage unit.

Reference numeral 307 denotes a speech recognition dictionary storage unit for storing a dictionary used for speech recognition, and reference numeral 308 denotes a dialogue control unit 3.
The dictionary selection unit 309 selects a dictionary to be used for speech recognition from the dictionaries stored in the speech recognition dictionary storage unit 307 in accordance with the instruction of 05, and the message stored in the message dictionary storage unit 310 in accordance with the instruction of the dialog control unit 305. Is a message selection unit for selecting a message to be presented to the user by voice from the list, and 310 is a message dictionary storage unit for storing a message to be presented to the user.

[0008] Next, the operation of the above-described conventional voice interactive device will be described with reference to FIGS. 7 and 8. The flow of the dialog shown below refers to FIG. 8, and the contents of the dictionary used for the dialog of the voice recognition refer to FIGS. 4 to 6. FIG. 4 is a diagram showing the contents of a speech recognition dictionary for speech recognition of a genre of a search item in the speech dialogue device, and FIG. 5 is the contents of a speech recognition dictionary for speech recognition of a prefecture with a golf course in the speech dialogue device. FIG. 6 is a diagram showing the contents of a voice recognition dictionary for voice recognition of a golf course in Shizuoka Prefecture by the voice dialogue device.

First, when a voice dialogue is started according to a user's instruction, a dialogue control unit 305 instructs a dictionary selection unit 308 to create a dictionary composed of words representing a search genre. In response to this command, the dictionary selection unit 308 creates a speech recognition dictionary composed of words representing the genre of the search from the speech recognition dictionary storage unit 307, as shown in FIG.

Next, the dialogue control unit 305 instructs the message selection unit 309 to output a message prompting the user to speak a word indicating the type of facility. In response to this command, the message selection unit 309 selects a message “Please select the type of facility” from the message dictionary storage unit 310 and presents it to the user by voice.

Next, the dialogue control unit 305 is
4 instructs to execute speech recognition using the dictionary created by the dictionary selection unit 308. The user who hears the message "Please select the type of facility" utters a word indicating a genre to be searched, in this case, "golf course" and inputs a voice signal to the voice interaction device. A characteristic parameter of the input speech signal is obtained by the acoustic analysis unit 303 and is recognized by the speech recognition unit 304.

As a result of the recognition, "golf course" is selected as a genre of search. The dialog control unit 305 stores this result. Next, the dialogue control unit 305 sets the dictionary selection unit 308
To create a dictionary composed of words representing the names of prefectures to be searched. In response to this command, the dictionary selection unit 308 creates a speech recognition dictionary composed of words representing the prefecture name to be searched, as shown in FIG. 5, from the speech recognition dictionary storage unit 307.

Next, the dialogue control unit 305 instructs the message selection unit 309 to output to the user the utterance of a word representing the name of the prefecture to be searched as a message. In response to this command, the message selection unit 309 selects a message "Please select the prefecture where the golf course is located" from the message dictionary storage unit 310 and presents it to the user by voice.

Next, the dialogue control unit 305 controls the speech recognition unit 3
When the user issues a command to execute voice recognition using the dictionary created by the dictionary selection unit 308, the user who hears the message “Please go to the name of the prefecture where the golf course is located” The word to be expressed, in this case, "Shizuoka Prefecture" is uttered and input to the voice interaction device. The input speech signal is determined by the acoustic analysis unit 303 for feature parameters, and is recognized by the speech recognition unit 304. As a result of the recognition, "Shizuoka prefecture" is selected as the name of the prefecture to be searched.

The result is stored in the dialogue control unit 305.
The dialogue control unit 305 determines “Shizuoka” as the result of the previous speech recognition.
And "golf course" which is the result of the speech recognition performed before that, and instructs the dictionary selection unit 308 to create a dictionary composed of the names of golf courses in Shizuoka Prefecture.
In response to this command, the dictionary selection unit 308 creates a speech recognition dictionary composed of the names of golf courses in Shizuoka Prefecture from the speech recognition dictionary storage unit 307 as shown in FIG.

Next, the dialogue control unit 305 instructs the message selection unit 309 to output to the user, as a message, the utterance of a word indicating the name of the golf course in Shizuoka Prefecture to be searched. In response to this command, the message selection unit 309 sends the message
The message "Please name the golf course" is selected and presented to the user by voice.

Next, the dialog control unit 305 controls the speech recognition unit 3
04, the dictionary selection unit 308 instructs to execute voice recognition using the dictionary created. When the user hears the message “Please name the golf course”, the user receives the name of the golf course to be searched. In this case, "○○
"Country club" and input it to the voice interaction device. The input speech signal is determined by the acoustic analysis unit 303 for feature parameters, and is recognized by the speech recognition unit 304. As a result of the recognition, "XX country club" is selected, and the search target is determined.

Next, the dialogue control unit 305 instructs the message selection unit 309 to present the determined search target “XX country club” to the user. In response to this command, the message selection unit 309 combines the content stored in the message dictionary storage unit 310 with "XX country club" to create a message "displays a map near the XX country club." And present it to the user by voice. Then, the map is displayed. With the above operation, the flow of the dialog shown in FIG. 8 is completed.

[0019]

However, since the above-mentioned widely used speech recognition apparatus does not have a means for storing a plurality of inputs, the speech recognition to be performed next is performed based on the previously input words. By narrowing down the target of, the method of searching for the desired item is adopted,
In the example of the golf course search described above, the flow of the dialogue is fixed to the one shown in FIG.

In general, in the field of voice interactive devices, the user of the voice interactive device does not feel uncomfortable or stress,
There is a need to provide natural spoken dialogue. In the above example, the name of the golf course is the subject of the information input by the user, and the prefecture name is supplementary information. For this reason, as shown in FIG. 8, when the supplementary information is input first and the main information is input later, the main information cannot be input first as compared to the opposite case. There is a problem that tends to give a sense of discomfort to the elderly.

The present invention has been made to solve the above-mentioned conventional problem. Since the number of recognized vocabulary words is limited by the performance limit of the speech recognition unit, first, supplementary information is input first to reduce the number of recognized vocabulary words. Even if it is necessary to enter the main information after narrowing down, use the voice recognition unit of the same performance to input the main information first and then input the supplementary information even if it is necessary to input the main information It is an object of the present invention to provide a spoken dialogue apparatus and a dialogue method that can realize the flow of the dialogue and can search the target item by changing the utterance order.

[0022]

SUMMARY OF THE INVENTION According to the present invention, there is provided a speech dialogue apparatus and dialogue method for accumulating an input speech signal in the form of an input speech signal or in the form of a characteristic parameter obtained by analyzing the input speech signal. And by recognizing the voice by changing the input order of the voice signals, the voice recognition result for the previously uttered word is narrowed down from the voice recognition result of the word uttered later.

According to the present invention, even when supplementary information is input after the main information is first input (uttered) using the voice recognition unit having the same performance, the speech recognition result of the word uttered later is obtained. By being able to narrow down the target of speech recognition for previously uttered words, a speech dialogue apparatus and dialogue method that does not give a user a sense of discomfort can be obtained.

[0024]

According to the first aspect of the present invention, switching between storing or analyzing an input voice signal and analyzing the stored input voice signal is performed according to a command from a dialogue control unit. Input voice control means, input voice storage means for storing an input voice signal according to a command of the dialogue control unit,
Acoustic analysis means for analyzing input speech signals to obtain feature parameters, and speech for performing speech recognition by comparing feature parameters obtained by analyzing an input speech signal in accordance with a command of a dialog control unit with a speech recognition dictionary. Recognition means, a dialogue control unit for controlling a voice dialogue, and a message selection means for selecting and outputting a message to be presented to the user from messages stored by a command of the dialogue control unit, The input voice signal is stored in the input voice storage means,
The voice recognition is performed by changing the order of the input voice signals, and by recognizing the voice by changing the order of the input voice signals, a voice interactive device that can change the voice order and search for a target item is realized. It has the effect of being obtained.

The invention according to claim 2 of the present invention is obtained by analyzing an input voice signal and analyzing the input voice signal in accordance with a command from a dialogue control unit, and acoustic analysis means for obtaining a characteristic parameter of the input voice signal. Parameter control means for switching between storing the feature parameter or performing voice recognition or recognizing the stored feature parameter as voice, and a feature parameter obtained by analyzing the input voice signal in accordance with a command from the dialog control unit Parameter storage means for storing voice data, voice recognition means for performing voice recognition by comparing a feature parameter obtained by analyzing an input voice signal in accordance with a command of the dialog control unit with a voice recognition dictionary, and dialogue for controlling voice dialogue A message to be presented to the user is selected from the messages stored in the message dictionary storage means according to a command from the control unit and the dialog control unit. Consists of a message selection means for outputting, to accumulate the characteristic parameters obtained by analyzing the audio signal input to the parameter storage unit, which has to be speech recognition out of sequence of feature parameters,
By reversing the order of the feature parameters of the input speech signal and performing speech recognition, an effect is obtained that a speech dialogue apparatus capable of changing the utterance order and searching for a target item is obtained.

According to the invention of claim 3 of the present invention, a speech signal by dialogue is inputted, the inputted speech signal is analyzed to obtain a characteristic parameter, and the characteristic parameter is obtained, and the speech signal is obtained from the inputted speech signal or from the speech signal. A feature parameter is accumulated, a message to be presented is selected and presented from messages stored according to the flow of the dialogue under the control of the control means, and the sequence of the stored voice signal or the feature parameter is defined as the flow of the dialogue. The voice recognition is performed by comparing the input voice signal with the voice recognition dictionary differently, and the voice recognition sequence is changed by performing voice recognition by changing the sequence of the input voice signal or the feature parameter of the voice signal. This has the effect that a voice dialogue method capable of searching for a target item is obtained.

Hereinafter, based on the attached drawings and FIGS.
An embodiment of the present invention will be described in detail. FIG. 1 is a block diagram illustrating a configuration of a voice interaction device according to a first embodiment of the present invention, FIG. 2 is a block diagram illustrating a configuration of a voice interaction device according to a second embodiment of the present invention, and FIG. FIG. 4 is a flowchart showing a flow of a voice dialogue by the voice dialogue device shown in FIGS. 1 and 2; FIG. 4 is a diagram showing the contents of a voice recognition dictionary for voice recognition of a genre of a search item in the voice dialogue device; FIG. 6 is a diagram showing the contents of a voice recognition dictionary for voice recognition of a prefecture with a golf course in a voice interactive device, and FIG. 6 is a diagram showing the contents of a voice recognition dictionary for voice recognition of a golf course in Shizuoka with a voice interactive device. FIG.

(Embodiment 1) First, referring to FIG.
The configuration of the voice interaction device according to the first embodiment of the present invention will be described in detail. In FIG. 1, reference numeral 101 denotes whether to store an input voice signal according to a command from the dialog control unit 105,
An input audio control unit that switches between analyzing an input audio signal or analyzing a stored input audio signal;
Reference numeral 2 denotes an input voice storage unit that stores an input voice signal according to a command from the dialog control unit 105.

Reference numeral 103 denotes an acoustic analysis unit for analyzing input speech signals to obtain feature parameters, and 104 denotes a feature parameter obtained by analyzing an input speech signal and a speech recognition dictionary in accordance with a command from the dialog control unit 105. A voice recognition unit for performing voice recognition by collating with a voice; 105, a dialogue control unit for controlling a voice dialogue; 106, dialogue control information for storing information on a flow of the voice dialogue determined according to a user operation or a result of voice recognition A storage unit 107 is a speech recognition dictionary storage unit that stores a dictionary used for speech recognition.

Reference numeral 108 denotes a dictionary selection unit for selecting a dictionary used for speech recognition from the dictionaries stored in the speech recognition dictionary storage unit 107 in accordance with a command from the dialogue control unit 105.
Reference numeral 9 denotes a message selection unit for selecting a message to be presented to the user from among messages stored in the message dictionary storage unit 110 in accordance with a command from the dialog control unit 105. Reference numeral 110 denotes a voice presentation to the user. This is a message dictionary storage unit for storing messages to be processed.

The input voice control unit 101, the input voice storage unit 102, the sound analysis unit 103, the voice recognition unit 104, the dialog control unit 105, and the message selection unit 109 are respectively an input voice control unit, an input voice storage unit, and a sound analysis unit. Means, voice recognition means, dialog control means, and message selection means.

Next, referring to FIG. 1 and FIGS. 3 to 6,
The operation of the spoken dialogue apparatus according to the first embodiment of the present invention will be described in detail with reference to the dialogue flow shown in FIG. 3 as an example. First, when a voice dialogue is started according to a user's instruction, the dialogue control unit 105 instructs the dictionary selection unit 108 to create a dictionary composed of words representing a search genre. In response to this command, the dictionary selection unit 108 creates a speech recognition dictionary composed of words representing the genre of the search from the speech recognition dictionary storage unit 107, as shown in FIG.

Next, the dialog control unit 105 instructs the message selection unit 109 to output a message prompting the user to utter a word indicating the type of facility. In response to this command, the message selection unit 109 reads “what genre to search?” From the message dictionary storage unit 110.
Is selected and presented to the user by voice.
(Note that this presentation to the user is not only made by voice, but can also be made with a display on a display device, the same applies hereinafter).

Next, the dialog control unit 105 controls the speech recognition unit 1
04 instructs to execute speech recognition using the dictionary created by the dictionary selection unit 108, and instructs the input speech control unit 101 to output an input speech signal to the acoustic analysis unit. The user who hears the message "Which genre do you want to search for?" Speaks a word indicating the genre desired to be searched, in this case, "golf course" and inputs it to the voice interaction device. The input speech signal passes through an input speech control unit 101, a characteristic parameter of which is obtained by an acoustic analysis unit 103, and is recognized by a speech recognition unit 104.

As a result of the recognition, "golf course" is selected as a genre of search. The dialog control unit 105 stores this result. Next, the dialogue control unit 105 instructs the message selection unit 109 to output the result of the previous voice recognition, “golf course”.
Is output as a message that prompts the user to utter the name. In response to this command, the message selection unit 109 selects a message “What golf course?” From the message dictionary storage unit 110 and presents it to the user by voice.

Next, the dialogue control unit 105 instructs the input voice control unit 101 and the input voice storage unit 102 to store the input voice signal. The input voice control unit 101 outputs the voice signal input according to this command to the voice storage unit 102, and the input voice storage unit 102 starts storing the input voice signal.

When the user hears the message "What is the golf course?" Presented earlier, the user utters a word indicating the golf course desired to be searched, in this case, "XX country club", and sounds. Type in the interactive device. The input voice “XX country club” is stored in the input voice storage unit 102 via the input voice control unit 101.

When the accumulation is completed, the dialog control unit 105
Commands the dictionary selection unit 108 to create a dictionary composed of words representing prefecture names to be searched. In response to this command, the dictionary selection unit 108 creates a speech recognition dictionary composed of words representing the name of the prefecture to be searched, as shown in FIG. 5, from the speech recognition dictionary storage unit 107.

Next, the dialogue control unit 105 instructs the message selection unit 109 to output to the user, as a message, a word that prompts the user to utter a word indicating the name of the prefecture to be searched. In response to this command, the message selection unit 10
9 selects a message "Which prefecture is it?" From the message dictionary storage unit 110 and presents it to the user by voice.

Next, the dialog control unit 105 controls the speech recognition unit 1
04 to execute voice recognition using the dictionary created by the dictionary selection unit 108, and to the input voice control unit 101,
Command to output to. Upon hearing the message "Which prefecture is it?", The user utters a word representing the prefecture to be searched, in this case, "Shizuoka prefecture" and inputs it to the voice interaction device. The input speech signal “Shizuoka” passes through the input speech control unit 101, a feature parameter is obtained by the acoustic analysis unit 103, and the feature parameter is recognized by the speech recognition unit 104. As a result of the recognition, “Shizuoka” is searched. Selected as the target prefecture name.

The dialogue control unit 105 stores the result and combines the result of speech recognition "Shizuoka" with the result of speech recognition "Golf course" performed before that to select a dictionary. It instructs the unit 108 to create a dictionary composed of the names of golf courses in Shizuoka Prefecture. In response to this command, the dictionary selection unit 108 creates a speech recognition dictionary composed of the names of golf courses in Shizuoka Prefecture from the speech recognition dictionary storage unit 107, as shown in FIG.

Next, the dialogue control unit 105 sends the voice signal of “XX Country Club”, which is the voice of the user previously stored, to the sound analysis unit 103 to the input voice control unit 101 and the input voice storage unit 102. Command to output to. In response to this command, the voice storage unit 102 outputs the stored voice signal to the input voice control unit 101, and the input voice control unit 101
Starts outputting the input audio signal to the acoustic analysis unit 103. The speech signal is analyzed by the acoustic analysis unit 103 to determine its characteristic parameters.
Recognized by From the recognition result, “XX Country Club” as shown in FIG. 6 is selected and the search target is determined.

Next, the dialogue control unit 105 instructs the message selection unit 109 to present the determined search target "XX country club" to the user by voice. In response to this command, the message selection unit 109 combines the content stored in the message dictionary storage unit 110 with "XX country club" to create a message "displays a map near the XX country club." To the user by voice. Then, the map is displayed. With the above operations, the flow of the dialog shown in FIG. 3 is completed.

(Embodiment 2) Next, referring to FIG.
The configuration of the voice interaction device according to the second embodiment of the present invention will be described in detail. In FIG. 2, reference numeral 201 denotes an acoustic analysis unit that analyzes an input audio signal and obtains a characteristic parameter thereof, and 202 stores a characteristic parameter obtained as a result of analyzing the input audio signal in accordance with a command from the dialog control unit 205. Or a parameter control unit that switches between speech recognition of a feature parameter obtained as a result of analyzing an input speech signal and speech recognition of a stored feature parameter.

Numeral 203 denotes a parameter storage unit for accumulating characteristic parameters obtained by analyzing the input voice signal in accordance with a command from the dialog control unit 205, and 204 a dialog control unit 20.
5, a voice recognition unit for performing voice recognition by comparing a feature parameter obtained by analyzing an input voice signal with a voice recognition dictionary, 205 is a dialogue control unit for controlling a voice dialogue,
Reference numeral 206 denotes a dialog control information storage unit that stores information on a flow of a voice dialogue performed according to a user operation or a result of voice recognition. Reference numeral 207 denotes a voice recognition dictionary storage unit that stores a dictionary used for voice recognition.

Reference numeral 208 denotes a command from the dialog control unit.
A dictionary selection unit 209 for selecting a dictionary to be used for speech recognition from the dictionaries stored in the speech recognition dictionary storage unit. A message selection unit 210 selects a message to be presented to the user, and a message dictionary storage unit 210 stores a message to be presented to the user.

Incidentally, the acoustic analysis unit 201, the parameter control unit 202, the parameter storage unit 203, the speech recognition unit 204,
The dialog control unit 205 and the message selection unit 209 correspond to an acoustic analysis unit, a parameter control unit, a parameter storage unit, a voice recognition unit, a dialog control unit, and a message selection unit, respectively.

Next, referring to FIG. 2 and FIGS. 3 to 6,
The operation of the spoken dialogue device according to the second embodiment of the present invention will be described in detail with reference to the flow of the dialogue shown in FIG. 3 as an example. First, when a voice dialogue is started according to a user's instruction, the dialogue control unit 205 instructs the dictionary selection unit 208 to create a dictionary composed of words indicating a genre of search. In response to this command, the dictionary selection unit 208 creates a speech recognition dictionary composed of words indicating the genre of search from the speech recognition dictionary storage unit 207 as shown in FIG.

Next, the dialogue control unit 205 instructs the message selection unit 209 to output a message prompting the user to speak a word indicating the type of facility. In response to this command, the message selection unit 209 reads from the message dictionary storage unit 210 “What genre to search?”
Is selected and presented to the user by voice.

Therefore, the dialog control unit 205 instructs the speech recognition unit 204 to execute speech recognition using the dictionary created by the dictionary selection unit 208, and also instructs the parameter control unit 202 to use the acoustic analysis unit. In step 201, an instruction is issued to output the characteristic parameters obtained as a result of analyzing the input voice signal to the voice recognition unit 204. The user who hears the message "Which genre do you want to search for?" Speaks a word indicating the genre desired to be searched, in this case, "golf course" and inputs it to the voice interaction device.

The input voice signal “golf course” is analyzed by the acoustic analysis unit 201 and converted into characteristic parameters.
Through the parameter control unit 202, the feature parameters obtained by the sound analysis unit 201 are recognized by the speech recognition unit 204. “Golf course” is selected as a search genre as a recognition result. This result is stored in the dialog control unit 205. Next, the dialog control unit 205 sets the message selection unit 209
Is instructed to output a message prompting the user to utter the name of the golf course, which is the result of the previous speech recognition. In response to this command, the message selection unit 209 selects a message “What golf course?” From the message dictionary storage unit 210 and presents it to the user by voice.

Next, the dialog control unit 205 instructs the parameter control unit 202 and the parameter storage unit 203 to store the characteristic parameters obtained as a result of analyzing the input voice signal in the sound analysis unit 201. In response to this command, the parameter control unit 202 outputs the input audio signal to the parameter storage unit 203, and the input parameter storage unit 203
Starts accumulation of feature parameters obtained by analyzing the input speech signal.

When the user hears the message "What is the golf course?" Presented earlier, the user speaks a word indicating the golf course desired to be searched, in this case, "XX country club". Input to the spoken dialogue device. The input audio signal "XX Country Club" is the sound analysis unit 2.
01, is converted into characteristic parameters, and is passed through the parameter control unit 202 to the parameter storage unit 2.
03.

When the accumulation is completed, the dialog control unit 205
Commands the dictionary selection unit 208 to create a dictionary composed of words representing the names of prefectures to be searched. With this directive,
The dictionary selection unit 208 receives the data from the speech recognition dictionary storage unit 207 as shown in FIG.
A speech recognition dictionary composed of words indicating the name of the prefecture to be searched is created as shown in FIG.

Next, the dialog control unit 205 instructs the message selection unit 209 to output to the user, as a message, a word that prompts the user to utter a word indicating the name of the prefecture to be searched. In response to this command, the message selector 20
9 selects a message "Which prefecture is it?" From the message dictionary storage unit 210 and presents it to the user by voice.

Next, the dialog control unit 205 controls the speech recognition unit 2
04 to execute voice recognition using the dictionary created by the dictionary selection unit 208, and instruct the parameter control unit 202 to input the input voice signal to the acoustic analysis unit 20.
The characteristic parameters obtained by the analysis in step 1
4 is output.

The user who hears the message "Which prefecture are you located in?" Speaks the word representing the prefecture to be searched, in this case, "Shizuoka prefecture" and inputs it to the voice interactive device. The input audio signal "Shizuoka" is the sound analysis unit 2.
01 is converted into characteristic parameters, and the characteristic parameters obtained by the acoustic analysis unit 201 are recognized by the speech recognition unit 204 via the parameter control unit 202.

As the recognition result, "Shizuoka prefecture" is selected as the name of the prefecture to be searched. This result is stored in the dialog control unit 205. The dialogue control unit 205 combines “Shizuoka Prefecture” as the result of the previous voice recognition with “Golf course” as the result of the previous voice recognition, and instructs the dictionary selection unit 208 to use the golf course in Shizuoka Prefecture. Command to create a dictionary composed of the names of In response to this command, the dictionary selection unit 208 creates a speech recognition dictionary composed of the names of golf courses in Shizuoka Prefecture from the speech recognition dictionary storage unit 207 as shown in FIG.

Next, the dialogue control unit 205 outputs the previously stored characteristic parameters of “XX country club” by the user's voice to the speech recognition unit 204 to the parameter control unit 202 and the parameter storage unit 203. Command. With this command, the parameter storage unit 203
The stored characteristic parameters are output to the parameter control unit 202, and the parameter control unit 202 starts outputting the characteristic parameters to the speech recognition unit 204. This feature parameter is recognized by the voice recognition unit 204, and as a result of the recognition, “OO country club” is selected and the search target is determined.

Next, the dialogue control unit 205 instructs the message selection unit 209 to present the determined search target “XX country club” to the user. In response to this command, the message selection unit 209 combines the content stored in the message dictionary storage unit 210 with "XX country club" to create a message "Display a map near the XX country club." To the user by voice. Then, the map is displayed. With the above operations, the flow of the dialog shown in FIG. 3 is completed.

[0061]

The present invention is configured as described above, and in particular,
An input voice storage unit for temporarily storing an input voice signal or a parameter storage unit for temporarily storing feature parameters as a result of analyzing the input voice signal is provided. By enabling speech recognition by narrowing down the vocabulary to be recognized, it is possible to realize a dialog flow in which the main information is input first, and then the supplementary information is input, using the same performance of the speech recognition unit. It is possible to provide a user-friendly voice dialogue apparatus and a dialogue method that can change the utterance order and search for a target item.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a configuration of a voice interaction device according to a first embodiment of the present invention.

FIG. 2 is a block diagram showing a configuration of a voice interaction device according to a second embodiment of the present invention;

FIG. 3 is a flowchart showing a flow of a voice dialogue by the voice dialogue device shown in FIGS. 1 and 2;

FIG. 4 is a diagram showing contents of a voice recognition dictionary for voice recognition of a genre of a search item in a voice interactive device.

FIG. 5 is a diagram showing contents of a voice recognition dictionary for voice recognition of a prefecture name of a golf course in the voice dialogue device.

FIG. 6 is a diagram showing the contents of a voice recognition dictionary for voice recognition of a golf course in Shizuoka prefecture by a voice dialogue device.

FIG. 7 is a block diagram showing a configuration of a conventional voice interaction device.

8 is a flowchart showing a flow of a voice dialogue by the voice dialogue device shown in FIG. 7;

[Explanation of symbols]

 101 input voice control unit 102 input voice storage unit 103 sound analysis unit 104 voice recognition unit 105 dialog control unit 106 dialog control information storage unit 107 voice recognition dictionary storage unit 108 dictionary selection unit 109 message selection unit 110 message dictionary storage unit 201 sound Analysis unit 202 Parameter control unit 203 Parameter storage unit 204 Voice recognition unit 205 Dialogue control unit 206 Dialogue control information storage unit 207 Voice recognition dictionary storage unit 208 Dictionary selection unit 209 Message selection unit 210 Message dictionary storage unit 303 Sound analysis unit 304 Voice Recognition unit 305 Dialog control unit 306 Dialog control information storage unit 307 Voice recognition dictionary storage unit 308 Dictionary selection unit 309 Message selection unit 310 Message dictionary storage unit

Claims (3)

[Claims] 1. An input voice control means for switching between accumulating or analyzing an input voice signal or analyzing a stored input voice signal in accordance with a command from the dialog control unit, and an input voice signal in response to a command from the dialog control unit. Input voice storage means for storing signals, acoustic analysis means for analyzing input voice signals to obtain feature parameters, feature parameters and voice recognition dictionary obtained by analyzing input voice signals in accordance with commands from a dialogue control unit A voice recognition means for performing voice recognition by collating with a user, a dialogue control unit for controlling a voice dialogue, and selecting a message to be presented to the user from messages stored by a command of the dialogue control unit. Means for selecting and outputting the input voice signal, storing the input voice signal in the input voice storage means, and changing the order of the input voice signal to perform voice recognition. Spoken dialogue apparatus according to symptoms. 2. An acoustic analysis means for analyzing an input voice signal to obtain a characteristic parameter thereof, and storing or recognizing a voice parameter obtained by analyzing the input voice signal according to a command of a dialogue control unit. A parameter control means for switching between voice recognition of the stored feature parameters and a parameter storage means for storing feature parameters obtained by analyzing an input voice signal in accordance with a command from the dialog control unit; A voice recognition unit for performing voice recognition by comparing a feature parameter obtained by analyzing an input voice signal with a voice recognition dictionary according to a command of a unit, a dialog control unit for controlling a voice dialog, and a command of the dialog control unit. Message selection means for selecting and outputting a message to be presented to the user from messages stored in the message dictionary storage means It accumulates the characteristic parameters obtained by analyzing the audio signal input to the parameter storage unit, the speech dialogue system, characterized in that so as to speech recognition out of sequence of feature parameters. 3. An audio signal by dialogue is input, the input audio signal is analyzed to determine a characteristic parameter, and the input audio signal or the characteristic parameter obtained from the audio signal is stored, and control of the control means is performed. A message to be presented is selected from the stored messages according to the flow of the dialogue and presented, and the stored voice signal or the order of the feature parameter is changed so as to be different from the flow of the dialogue, and is compared with the voice recognition dictionary. A voice interaction method comprising the steps of: