JPH0916191A - Device and method for speech recognition for navigator - Google Patents

Abstract

PURPOSE: To shorten the speech recognition time of a destination which is inputted in voice. CONSTITUTION: A standard feature parameter file which is used for speech recognition is divided into rough sections - small sections and a speech is divided and inputted by the sections. The file is used for each section to perform the speech recognition by an MPU 13. Recognition candidates whose similarities are larger than a threshold value are displayed on a display device 17 and a user selects a desired recognition candidate on an operation panel 18.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a voice recognition device and method for a navigator which is mounted on an automobile and is suitable for a navigator which guides a route from a local point to a destination.

[0002]

2. Description of the Related Art Navigators are installed in automobiles and are rapidly becoming popular as devices for guiding a route from a local point to a destination. In the initial navigator, the driver operated the operation panel to manually input the destination of the route search target. When a driving driver operates the operation panel of the navigator, it often takes time to search and register the destination, and a navigator that can input the destination by voice is proposed to improve the operability. Has been done. Conventionally, in this type of navigator, when the destination is input by voice,
Voice recognition is executed by the built-in voice recognition device, and a character string indicating the destination is acquired. This character string is displayed on the display device. In order to instruct the navigator whether the displayed character string is the destination intended by the driver, that is, whether the recognition result is correct or incorrect, the speaker does not speak for a certain period of time. By doing so, the correct answer is instructed, and by uttering some word, the incorrect answer is instructed.

Further, by introducing a dialogue system into the voice recognition device and inputting a word such as "yes" or "no" to the voice recognition device with respect to a single recognition result output from the voice recognition device, the correct answer / A voice recognition device for navigator has also been proposed in which an incorrect answer is instructed to obtain a finally confirmed final recognized word.

[0004]

When the map to be searched for a route is spread over a wide area such as a national map, a plurality of same-sound place names are likely to occur and erroneous recognition is likely to occur. For this reason, there is a problem that it takes time for the final recognition result to determine the word.

Therefore, in view of the above points, the present invention is to provide a voice recognition device and method for a navigator capable of shortening the time until the recognition result is determined.

[0006]

In order to achieve such an object, the invention of claim 1 inputs a place name to be searched for a route by voice, and extracts a characteristic parameter extracted from the input voice. And the standard feature parameter stored in association with the place name are compared with each other, and the standard feature parameter stored in the navigator voice recognition device that uses the place name corresponding to the standard feature parameter having a high degree of similarity as the voice recognition result. And corresponding place names are divided into regions and the place names input by voice are determined and stored in the order of the regions, and the standard feature parameters and the corresponding place names whose similarity is equal to or greater than a threshold value. Detecting means for detecting the detected place names, information processing means for rearranging the detected place names in the order of similarity, display means for displaying a plurality of the rearranged place names, and the display Characterized in that comprises an indication means for indicating the determined to be the place name of a plurality of place names that.

According to a second aspect of the present invention, a place name to be searched for a route is input by voice, and a characteristic parameter extracted from the input voice and a standard feature parameter stored in association with the place name. In the voice recognition method of the voice recognition device for navigator in which the place name corresponding to the standard feature parameter having a high degree of similarity is the voice recognition result,
The standard feature parameters and corresponding place names to be stored are
The area names are divided and stored for each area division, and the place names input by voice are set in the area division order, and the place names are detected for each area division, and the standard feature parameters and the similarity are equal to or more than a threshold value. Detects the corresponding place name, rearranges the detected place names in the order of their similarity, displays a plurality of the rearranged place names, and indicates the place name to be confirmed among the plurality of displayed place names, The feature is that the place names are determined in the order of division.

[0008]

According to the invention of claims 1 and 2, instead of the conventional example in which the recognition methods are displayed one by one, recognition candidates having a predetermined degree of similarity or higher are displayed, and the speech recognition processing time is reduced by selecting from the candidates. To shorten.

According to the first and second aspects of the present invention, the place name is voice-recognized for each segment, the recognition candidates having a predetermined degree of similarity or higher are displayed, and one of the candidates is selected, so that the search time of the standard feature parameter is shortened. Therefore, the final correct recognition result can be obtained in a short time.

[0010]

Embodiments of the present invention will be described below in detail with reference to the drawings.

FIG. 1 shows a system configuration of a navigator to which the present invention is applied, showing a system configuration related to a built-in voice recognition device. In FIG. 1, a voice input from the microphone 20 for registration of a destination is converted into an analog electric signal and converted into a digital signal by the A / D converter 11. This digital signal is transmitted via the input / output interface (I / O) 12 to the microprocessor (MP
U) 13 is input.

The MPU 13 extracts a voice characteristic parameter from the input digital voice signal in a conventional manner. In the CD (compact disc) in the CD drive 15, standard characteristic parameters known in advance for each place name are stored in association with the place name. In this embodiment, as shown in FIG. 2, (prefecture), (city,
District names (towns, villages) are divided into large, medium, and small categories, and place names, that is, recognition words corresponding to standard feature parameters are prepared.

The CD loaded in the CD drive 15 has
Needless to say, the location information of the place name necessary for the route search is also stored. Returning to FIG. 1, the MPU 13
Is a plurality of standard feature parameters similar to the feature parameters acquired from the input voice signal,
The candidate of the recognition word (character code form) corresponding to the detected standard feature parameter is detected from the above and is read out via the I / O 14.

The MPU 13 converts the acquired recognition word candidate into a displayable image (font pattern) from the character code and delivers it to the display controller 16. The display controller 16 outputs this image information to the display device 17. As a result, a plurality of recognition word candidates are displayed on the display screen of the display device 17. The user operates the selection button on the operation panel 18 to select the recognition word candidate intended by the user. This selection result is confirmed as the voice recognition result.

The speech recognition processing of the present invention executed in such a system configuration will be described with reference to the flowchart of FIG. FIG. 3 shows the main steps of the voice recognition processing for the destination. In this embodiment, the destination names generated by the user are predetermined in the order of prefecture name → city, ward → town, village.

The user: 1) As a prefecture name, for example, "Kanagawaken" is input to the microphone 20 by voice.

2) As described above, the MPU 13 extracts the characteristic parameters from the input voice signal, searches the standard parameter file for the prefectures on the CD (see FIG. 2), and selects candidate words in descending order of similarity. Extract and MPU13
Temporarily store in internal memory. More specifically, the standard feature parameter at the head of the standard feature parameter file of the prefecture on the CD is read and the distance to the feature parameter extracted from the audio signal is calculated. When the distance calculation result is within the allowable range (high similarity), the recognition word of the standard feature parameter and the distance calculation result are stored in the memory in the MPU 13.

Hereinafter, the standard feature parameters are sequentially read from the standard feature parameter files of prefectures on the CD,
Calculate distances and compare thresholds. When a standard feature parameter whose distance is within the allowable range is found, the size relationship with all the data of the distance calculation results stored in the memory is checked, and the data in the memory in order of similarity (record consisting of recognition word and distance) Are rearranged (step S in FIG. 4)
100-S120). When such a process is repeated and executed and the above process is performed for all standard feature parameters of all prefectures, the recognition words are stored in the memory in the MPU 13 in descending order of similarity (step S1 in FIG. 3).
0).

When the recognized word is displayed as shown in FIG. 5, the user designates a desired recognized word candidate with the number buttons on the operation panel 18. If there is no desired recognition word candidate, the user inputs the voice again.

The recognition word designated by the number buttons is
It is stored in the memory in the MPU 13 as a voice recognition result for all prefectures. The user then voice-inputs the city name. The MPU 13 sets a standard feature parameter file for city names on the CD as a search target, performs the same processing as described above to display all recognition candidates on the display screen, and accepts recognition word candidates designated by the user ( Steps S30 to S40 of FIG. 3). Similar processing is performed for the town name (steps S50 to S60 in FIG. 3). Based on the destination determined in this way, the MPU 13 performs route search processing.

As described above, in the present embodiment, since the recognition word candidates having the similarity of a predetermined value or more are displayed and the desired candidate among them is confirmed, a single recognition candidate is required as in the conventional case. The processing is simplified as compared with the determination method for.

The following example can be carried out in addition to this embodiment.

1) Each time the user decides a recognized word, the MPU 13 updates the frequency of use and stores the frequency of use in a non-volatile memory or a hard disk.

The homonyms are rearranged and displayed in the order of frequency of use. This method is effective when there are many places with the same sound, such as towns and villages.

2) In addition, frequently used place names are registered in advance in a memory separate from the CD file, using the voice or the operation panel to register the standard feature parameters and recognition words. When the destination of the route search target is input by voice, the memory is searched first, and then the CD file is searched. When such a search method is used, the process of recognizing frequently used place names becomes faster.

3) Further, in this embodiment, the learning function, that is, the confirmed recognition word and its standard feature parameter are stored in the memory, and the standard feature parameter is first referred to in the next speech recognition. The function may be added to the MPU 13.

4) In the present embodiment, an example in which the route is searched from the current position during traveling to the destination is shown, but the present invention can be applied to the case where the departure point is input by voice.

5) In the present embodiment, the operation panel 18 is used as an instruction means for selecting the displayed recognition word, but it is a matter of course that a voice input such as "ichi" or "ni" may be used. .

6) In the classification of place names, major categories such as "Kanto" and "Kinki" may be provided as high-ranking categories of prefectures. Also, a famous object name such as Tokyo Tower may be used instead of the village classification.

[0030]

As described above, according to the inventions of claims 1 and 2, by recognizing a place name by voice for each section, displaying recognition candidates having a predetermined similarity or higher, and selecting from among them,
The search time for the standard feature parameters becomes short, and the final correct recognition result can be obtained in a short time. .

Further, since the voice recognition processing time is shortened by these inventions, the danger of the driver in carrying out a route search while traveling is alleviated.

[Brief description of the drawings]

FIG. 1 is a block diagram illustrating a system configuration according to an embodiment of the present invention.

FIG. 2 is an explanatory diagram showing a file structure of standard feature parameters used for voice recognition.

FIG. 3 is a flowchart showing a voice recognition processing procedure of the present embodiment.

FIG. 4 is a flowchart showing a partial processing procedure of the voice recognition processing procedure of the present embodiment.

FIG. 5 is an explanatory diagram showing an example of a display according to the embodiment of the present invention.

[Explanation of symbols]

 10 Navigator 20 Microphone 11 A / D 12, 14, 15 I / O 13 MPU 15 CD Drive 17 Display Device 18 Operation Panel

Claims (2)

[Claims] 1. A place name to be searched for a route is input by voice, a feature parameter extracted from the input voice is compared with a standard feature parameter stored in association with the place name, and the similarity is calculated. In a voice recognition device for a navigator that uses a place name corresponding to a high standard feature parameter as a voice recognition result, the standard feature parameter and the corresponding place name to be stored are
A storage unit that divides each region into regions and stores the place names input by voice in the region segment order as well; and a detection unit that detects a standard feature parameter having a similarity not less than a threshold value and a corresponding place name. , An information processing means for rearranging the detected place names in the order of the degree of similarity, a display means for displaying a plurality of the rearranged place names, and an instruction for instructing a place name to be confirmed among the plurality of displayed place names. A voice recognition device for a navigator, characterized by comprising: 2. A place name to be searched for a route is input by voice, the feature parameter extracted from the input voice is compared with a standard feature parameter stored in association with the place name, and the similarity is calculated. In the voice recognition method of the navigator voice recognition device in which a place name corresponding to a standard feature parameter with a high degree is used as a voice recognition result, the standard feature parameter to be stored and the corresponding place name are
It is divided and stored for each area division, and the place names input by voice are also set in the area division order, the place names are detected for each area division, and the standard feature parameter and the similarity degree of which are equal to or more than a threshold value. The corresponding place name is detected, the detected place names are rearranged in the order of the degree of similarity, the plurality of rearranged place names are displayed, and the place name to be confirmed among the plurality of displayed place names is designated, A voice recognition method for a voice recognition device for a navigator, characterized in that place names are determined in order of division.