JP2000181485A - Device and method for voice recognition - Google Patents

Abstract

PROBLEM TO BE SOLVED: To lighten the burden of speaking on a user when a destination for navigation is vocally set. SOLUTION: The user inputs destination data for navigation through a microphone 10. A control part 14 stores the input spoken data in a spoken data storage part 22 and analyzes at least part of the input spoken data by using a voice data base 18. The voice data 18 used for the analysis is switched by using the analysis result and the spoken data stored in the spoken data storage part 22 are read out to take an analysis again. Thus, the stored spoken data are used, so the user speaks only once to set the destination with high precision.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a speech recognition apparatus and method, and more particularly to speech recognition when setting a destination in a navigation system.

[0002]

2. Description of the Related Art Hitherto, there has been proposed a technique for performing various processes of a navigation system by voice, for example, setting a destination. In such a technique, it is an important issue how to quickly and accurately recognize a voice uttered by a user. Normally, speech recognition is performed by comparing utterance data of a user with speech data in a speech database prepared in advance, but the speech database is often used in a hierarchical manner.

For example, Japanese Patent Application Laid-Open No. Hei 10-62199 discloses that an audio database is divided into three layers, a facility name having location information and a facility genre name having no location information are stored in layer 1, and a hierarchy 2 is stored in layer 2. A facility name having location information and a prefecture name having no location information corresponding to the genre name of 1 are stored, and a facility name having location information corresponding to the prefecture of the tier 2 is stored in the layer 3; A technology for recognizing speech by sequentially changing the hierarchy according to user utterance data is disclosed.

[0004]

However, in the above-mentioned prior art, since the hierarchy of the voice database is switched for each utterance of the user, even if the user wants to set the navigation destination to the department store ** department store, for example. , You must repeatedly say "facility" → "department store" → "** department store" repeatedly, and set the destination with one natural utterance like "I want to go to the department store ** department store" Could not.

[0005] In addition, some users may want to set a destination starting from a certain target, such as "a parking lot near a department store". There was a problem that the destination setting as the starting point could not be recognized.

SUMMARY OF THE INVENTION The present invention has been made in view of the above-mentioned problems of the related art, and has as its object to provide an apparatus and a method capable of reducing a user's utterance load and easily setting desired data by voice. Is to do.

[0007]

According to a first aspect of the present invention, an utterance data storage unit for storing utterance data of a user is compared with the utterance data and audio data in an audio database. First voice analysis means for analyzing at least a part of the utterance data, switching means for switching the voice database based on the analysis data obtained by the first voice analysis means, and storage in the utterance data storage means Second voice analysis means for re-analyzing the utterance data by reading out the generated utterance data and comparing it with the voice data in the voice database switched by the switching means. First
The analysis is performed by the voice analysis unit, and the stored utterance data is read out and re-analyzed by the second voice analysis unit (at the time of the re-analysis, the voice database is switched and optimized), so that the user's 1 Speech recognition can be reliably performed by utterance of the second time. It should be noted that the first sound analysis means and the second sound analysis means need not be provided separately, and both functions can be achieved by the same means.

According to a second aspect of the present invention, in the first aspect, the utterance data is destination data for navigation, and when predetermined data is obtained by the second voice analysis means, the data is converted to the destination data. It is characterized by having means for processing as the target for the destination. By taking predetermined data obtained by voice analysis as a target, for example, utterance data of "XX near **"
By obtaining the predetermined data “near **”, “*
By using "*" as a target, "OO" of the original destination can be obtained.

According to a third aspect of the present invention, at least a part of the voice data is analyzed by comparing the voice data of the user with voice data in a voice database by storing the voice data of the user and storing the voice data of the user. A first analysis step, a switching step of switching the voice database based on the analysis data obtained in the first analysis step, reading voice data stored in the storage step, and reading the read voice data and the switching. A second analysis step of reanalyzing by comparing with the audio data in the audio database switched in the step.

In a fourth aspect based on the third aspect, the utterance data is destination data for navigation, and when predetermined data is obtained in the second analysis step, the data is converted to the destination data. It is characterized by further comprising a processing step of processing as a target for a destination.

[0011]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, embodiments of the present invention will be described with reference to the drawings, taking a destination setting in a navigation system as an example.

FIG. 1 is a block diagram showing the configuration of the present embodiment, and shows the configuration of a navigation system having a voice recognition function.

The microphone 10 inputs a user's (vehicle occupant) 's utterance and supplies it to the control unit 14. Current position detector 12
Is composed of GPS, vehicle speed sensor, direction sensor, etc.
The current position of the vehicle is detected and supplied to the control unit 14.

The control unit 14 is specifically constituted by a microcomputer, executes various controls necessary for navigation, analyzes the user's utterance data input from the microphone 10, and sets a destination. In the present embodiment, the control unit 14 functions as a first voice analysis unit and a second voice analysis unit, and also functions as a switching unit that switches data used for analysis in the voice database 18.

The voice database 18 stores voice data to be compared with the voice data when the voice data of the user is analyzed by the control unit 14, and has a hierarchical structure. The control unit 14 appropriately accesses the voice database 18 and analyzes the voice data. Voice database 1
Reference numeral 8 includes, for example, a CD-ROM or a DVD.

The map data storage unit 20 stores map data (display map data and route search map data) necessary for navigation, and the control unit 14 stores map data around the detected current position in the map data. A route to a destination obtained by reading out from the storage unit 20 and displaying it on the display unit 24 or analyzing the utterance data is searched for using the route search map data, and is displayed as a recommended route.
To be displayed. Of course, the recommended route may be notified by voice from a speaker. The map data storage unit 20 stores, for example, a CD
-Consists of ROM, DVD, etc.

The utterance data storage unit 22 stores the utterance data input from the microphone 10, and the control unit 14 reads out the utterance data stored in the utterance data storage unit 22, so that the user can receive the utterance data again. Multiple analyzes of utterance data can be performed without requiring utterance. The utterance data storage unit 22 can be composed of, for example, a semiconductor memory.

The operation unit 16 is used for various input operations such as scrolling of map data displayed on the display unit 24 and manual destination setting without using voice.

FIG. 2 shows a hierarchical structure of the audio database 18. The speech database is composed of three layers: a nationwide recognition grammar dictionary, a prefectural level recognition grammar dictionary, and a municipal level recognition grammar dictionary. The “glamor dictionary” refers to the control unit 14
Is a set of voice data in the grammar technique used when analyzing the utterance data in the grammar technique, which will be described later. The national-level recognition grammar dictionary stores the data of major place names and names throughout Japan, and the prefectural-level recognition grammar dictionary stores the data of the place names and names within each prefecture, which are classified by prefecture. The data is stored in the recognition grammar dictionary at the municipal level, for each municipal level and the data of the place names and names in the municipal level.

The control section 14 determines which data of which layer in the voice database 18 is to be read and used in accordance with the detected current position and the analysis result of the utterance data. Specifically, for example, when the current position of the vehicle is Susono City, Shizuoka Prefecture, the control unit 14 designates Shizuoka Prefecture as the prefectural level recognition grammar dictionary in the voice database 18, and the municipal level recognition grammar dictionary. Is designated as Susono City and its neighboring cities. Then, when the vehicle moves and the current position of the vehicle becomes Chiyoda-ku, Tokyo, the control unit 14 designates Tokyo as the recognition grammar dictionary at the prefectural level and Chiyoda-ku as the recognition grammar dictionary at the municipal level. And the ward nearby. The advantage of specifying the voice database according to the current position is that when the voice is uttered around the current position as the destination, the voice database can be quickly analyzed and recognized. The control unit 14 switches the usage data of the voice database 18 according to the analysis result of the utterance data. For example, if the analysis of the utterance data indicates that Mishima City is the target, the municipal level recognition grammar dictionary is switched to Mishima City and analysis is continued.

FIG. 3 is a flowchart showing a speech recognition process according to this embodiment. First, the user speaks and inputs a destination (S101). As a form of the utterance, for example, "Mishimashi ** Shoten (*
* Shops) and I want to go to the parking lot near Mishima Station. These utterance data input from the microphone 10 are stored in the utterance data storage unit 22 and (S
102), the control unit 14 analyzes the input utterance data by using the grammar method (S103).

Here, the glamor technique will be described.
The grammar method is a method in which a sequence of words to be recognized is defined in advance and recognized, and for example, the sequence of sentences is defined as <a><b
><C>,<a> as a candidate for “today”, “tomorrow” or “tomorrow”, <b> as a candidate for “weather” or “weather”, and as <c>, “ Recognition of utterance data such as "the weather is good today", "the weather is bad today", "the weather is good tomorrow", and the like are set as good or bad candidates. In the present embodiment, the following word sequence (referred to as a phrase) is used to perform destination recognition.

Basic phrase 1 = <place name><end>? Basic phrase 2 = <NULL>? <Name><end>? Basic phrase 3 = <NULL> of <place name>? <Name><
end>? Basic phrase 4 = <NULL> of <name> of <place name>?
<Name><end>? Basic phrase 5 = <NULL> of <name>? <Name><
end>? Basic phrase 6 = <place name><direction><end>? Here, the <place name> is a word representing an address or an area, and the address is “Shizuoka” or “Shizuokaken”, and the area is “Izu” or “Busou”. Also, <end>
Is the word that indicates the end of the sentence, "I want to go,""I want to go,""I want to go,""I want to go,""Go,""Go,"
"Until", "Until only", "He", etc. <NULL>
Is a word that expresses the range or degree.
"Here". The <NULL> data is also data required when setting a target. <Name> is a word that indicates the name or facility, such as “** Eki”, “** Chushajo”, “** Gorfujo”, “** Koen”, or “**”.
"Inta Chang", "** Byoin", "** Minato"
"** kawa", "** kankochi", "** onsen" and the like. <Direction> is “** homen” or the like. Also, <
> After? Indicates that the word <> is not essential and need not be present. Therefore, basic phrase 1 contains
In addition to “I want to go to Shizuoka”, “Shizuoka” is also included. In the above example, "*
"* Shop (** shop in Mishima-shi)" corresponds to basic phrase 3, and "Mishima-eki-no-chikaku-no-chushojo (parking lot near Mishima station)" corresponds to basic phrase 5.

When the user's utterance data is analyzed using such a grammar technique, at least a part of the data can be analyzed, but the remaining data (particularly, data in the latter half of the utterance).
May not be able to be analyzed. Specifically, when the above-mentioned utterance data “Mishimashino ** Shoten (** shop in Mishima City)” is analyzed, “Mishimashi (Mishima City)” is recognized as a nationwide recognition grammar dictionary. Although it exists and can be analyzed, the name “** shoten (** shop)” cannot be analyzed unless it is a municipal level recognition grammar dictionary, and the municipal level recognition grammar dictionary is not available for municipalities other than Mishima. If it is specified (for example, if the current position of the vehicle is a supporting city, the default value of the municipal level recognition grammar dictionary is the supporting city), the utterance data cannot be analyzed. Therefore, the grammar dictionary of the voice database 18 is switched using the result obtained by the analysis (S104).
In the above case, since “Mishimashi (Mishima City)” is obtained, the recognition grammar dictionary at the municipal level is switched to data for Mishima City.

After switching the voice database 18, S
The utterance data stored in the utterance data storage unit 22 in the process of 102 is read and analyzed again (S105). At this time, since the recognition grammar dictionary at the municipal level is data for Mishima City, “** shoten (** store)” in the utterance data can be analyzed. When all the analysis of the utterance data is completed, the control unit 14 searches the map data for a destination using the analysis result (S106).
In this example, the map data for Mishima city is read and ** stores are searched.

On the other hand, if the utterance data is "I want to go to the parking lot near Mishima station", the analysis is performed in the same manner in S103, and the active recognition grammar is obtained. The dictionary (in this case, the national level) hits and can analyze "Mishima Eki (Mishima Station)", "Chikakuno (nearby)", "Chushajo (parking lot)". Then, the recognition grammar dictionary at the municipal level is switched to data for Mishima city (S104),
The utterance data stored in the utterance data storage unit 22 is read and analyzed again (S105). In this case, since all utterance data can be analyzed in the first analysis, the result of the second analysis is the same as the result of the first analysis. Of course, if the utterance data is "
* If you want to go to the chushajo (you want to go to the ** parking lot near Mishima station) ", in the first analysis, the part of" ** chushajo (** parking) "is analyzed. It becomes impossible, and the part of “**” can be analyzed in the second analysis after switching the voice database. Then, since “near”, which is <NULL> data, exists, the control unit 14 processes an analysis result of the <name> data existing before the <NULL> data as a target, and displays this target on the map data. The parking lot is searched in order from the coordinates (X, Y) of the object (Mishima Station) (S10).
6).

As described above, in the present embodiment, the utterance data of the user is stored, and even if the analysis cannot be performed in the first analysis, the voice database is automatically switched and the analysis is performed again. Is improved,
The user can set the destination with one utterance.

In this embodiment, when <NULL> data is present, the preceding data is regarded as a target and searched from the map data, and the map data around the target is searched for. Since the user can search for the destination, the user can easily set the desired destination with a natural utterance.

In the present embodiment, when a homonym exists, it is preferable to obtain more information from the user in order to improve the recognition rate. For example, when the user utters “Toyota”, the speaker inquires, “Is it OK?”.

Further, when analyzing the utterance data, a technique such as adding an annotation indicating the type of the obtained data to facilitate the search of the map database can be used in the present embodiment. For example,
The prefecture name in the place name is the number 11,
For example, the number 13 is assigned to the city name, the number 42 is assigned to the area, and the number 32 is assigned to the name as an annotation. In this case, a number is given as an annotation to “near” or “near” in the <NULL> data (for example, 91), while an annotation is given to “good” or “easy” in the <NULL> data. It is preferred not to apply. This is because these terms are unnecessary for setting the destination (searching the map data).

[0031]

As described above, according to the present invention, it is possible to reduce the user's utterance load and to more easily obtain desired data,
For example, a destination for navigation can be set by voice.

[Brief description of the drawings]

FIG. 1 is a configuration block diagram of an embodiment.

FIG. 2 is an explanatory diagram illustrating a structure of a voice database according to the embodiment;

FIG. 3 is a processing flowchart of the embodiment.

[Explanation of symbols]

10 microphones, 12 current position detection unit, 14 control unit,
16 operation unit, 18 voice database, 20 map data storage unit, 22 utterance data storage unit, 24 display unit.

──────────────────────────────────────────────────続 き Continued on the front page (51) Int.Cl. ⁷ Identification symbol FI Theme coat ゛ (Reference) G09B 29/10 G09B 29/10 Z F term (Reference) 2C032 HB06 HC16 HD16 2F029 AA02 AB01 AB07 AB09 AC01 AC02 AC04 AC18 5D015 HH13 HH16 KK02 LL10 5H180 AA01 BB04 BB13 CC12 CC27 FF04 FF05 FF22 FF25 FF32 FF33

Claims (4)

[Claims] An utterance data storage unit for storing utterance data of a user; a first voice analysis unit for analyzing at least a part of the utterance data by comparing the utterance data with voice data in a voice database; A switching unit that switches the voice database based on the analysis data obtained by the first voice analysis unit; and reads the utterance data stored in the utterance data storage unit, and reads the utterance data in the voice database that is switched by the switching unit. And a second voice analysis means for re-analyzing the voice data by comparing the voice data with voice data. 2. The apparatus according to claim 1, wherein the utterance data is destination data for navigation, and when predetermined data is obtained by the second voice analysis means, the data is converted to the destination data. A speech recognition device having means for processing as a target. 3. A storage step for storing utterance data of the user; a first analysis step of analyzing at least a part of the voice data by comparing the utterance data of the user with voice data in a voice database; A switching step of switching the voice database based on the analysis data obtained in the first analysis step; and reading the utterance data stored in the storage step,
A second analysis step of reanalyzing by comparing the read utterance data with the voice data in the voice database switched in the switching step. 4. The method according to claim 3, wherein the utterance data is destination data for navigation, and when predetermined data is obtained in the second analysis step, the data is converted to the target data for the destination. A speech recognition method, further comprising a processing step of processing as a thing.