JP2003241789A - Device and method for speech recognition dictionary creation - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To make a system not misrecognize a track number which is never possible to exist when a speech recognition is carried out with a track number of a CD. <P>SOLUTION: This device is provided with an attribute information input part 15 which acquires attribute information (e.g. the number of tracks) regarding the recording contents of the CD from an audio unit 20 controlled through speech recognition and a dictionary creation part 16 which creates a speech recognition dictionary 12 within a range of the number of commands corresponding to the acquired attribute information; and the speech recognition dictionary 12 having commands as many as pieces of music actually recorded on the CD as an object of speech recognition is dynamically created to exclude track numbers in advance, which are not possible to exist on the inserted CD including even track numbers which are possibly misrecognized as actually present track numbers, thereby eliminating the trouble that a track number which is not present on the CD is recognized by mistake. <P>COPYRIGHT: (C)2003,JPO

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an apparatus and method for creating a voice recognition dictionary, and more particularly, it is suitable for use in an apparatus and method for creating a voice recognition dictionary to be used when electronic equipment is controlled by voice recognition. It is a thing.

[0002]

2. Description of the Related Art Most of recent vehicles are equipped with various electronic devices. The electronic equipment installed, for example,
CD (compact disc), MD (mini disc),
The driver can easily reach the desired destination by playing the audio device that plays back the music recorded on the cassette tape or playing back the received radio broadcast, the air conditioner (air conditioner), and the running guide of the vehicle. There is a navigation device that can be used.

There is also an integrated system in which various electronic devices such as the audio device, the air conditioner, and the navigation device are connected by a bus. In such an integrated system, one resource can be shared by a plurality of electronic devices by exchanging information between the electronic devices via a bus. For example, one display device is shared by a plurality of electronic devices, and an audio screen, an air conditioner operation screen, a navigation screen, and the like can be appropriately switched and displayed.

Recently, in order to avoid a one-handed operation or the like when controlling an electronic device, a system has been provided in which the electronic device can be controlled by voice recognition.
By using this voice recognition technology, the driver can control various electronic devices without releasing his / her hand from the steering wheel (without manually operating an operation unit such as a remote controller or an operation panel). When controlling an electronic device by voice recognition, a voice recognition dictionary (database that collects recognition target vocabulary) for recognizing a specific control command in comparison with voice input from a microphone or the like is essential.

For example, as shown in FIG. 4, a voice recognition dictionary for voice recognition by designating a CD track number is as shown in FIG.
The association between the audio command such as “Seedee <track number>” and the command ID corresponding to each audio command is stored. When the user inputs a voice through a microphone or the like, the voice recognition dictionary in FIG. 4 is referred to, the audio command closest to the input voice is recognized, and the corresponding command ID is output. The command ID is supplied to the audio device, and the music of the designated track number is reproduced.

[0006]

However, in the case of performing voice recognition by the track number of the CD as described above,
Some track numbers sound similar to each other, and the system could erroneously recognize them. For example, when a track number is input by voice in English, for example, "fifteen" and "fifty" have very similar pronunciations and are confusing, and there is a high possibility of erroneous recognition.

By the way, the track numbers of the CD are 1 to 99.
Is defined in the standard. Therefore, in the voice recognition dictionary for CD as shown in FIG. 4, commands corresponding to track numbers 1 to 99 are prepared. However, in the case of a music CD, for example, the number of tracks actually recorded is at most 20 or less. Therefore, if voice recognition is performed using a voice recognition dictionary having track numbers up to 99, a track number that does not actually exist will be erroneously recognized (for example, the user pronounced “fifteen” but actually exists). Do not recognize "fifty"). If a track number that does not actually exist is erroneously recognized, the audio device cannot accept the command, and the usability for the user becomes poor.

There is also a voice recognition dictionary created by predicting the maximum number of tracks recorded on a CD and narrowing it down in advance to the range of the assumed maximum number (for example, 30 tracks). However, when using such a voice recognition dictionary,
For a CD having more tracks than the assumed maximum number, there is a problem that it is impossible to specify a track number larger than the assumed maximum number.

The present invention has been made in order to solve such a problem, and when the voice is recognized by the track number of the CD, the system does not mistakenly recognize the track number which should not exist. The purpose is to do. It is another object of the present invention to prevent the system from erroneously recognizing information that should not be present in the medium targeted for voice recognition, not limited to the CD track number.

[0010]

In order to solve the above-mentioned problems, according to the present invention, attribute information relating to recorded information of a medium used in an electronic device is controlled by an electronic device which is controlled by voice recognition. (For example, the number of pieces of record information, the title of each piece of record information, the file name of each piece of record information, etc.) is acquired, and a voice recognition dictionary is created within a range of the number of commands according to the acquired attribute information. According to the present invention having such a configuration, a voice recognition dictionary having as many commands as the number of pieces of information actually recorded in the medium to be voice-recognized is dynamically created. The voice recognition process will be performed using this. As a result, information that may be erroneously recognized can be removed from the voice recognition dictionary in advance, and the inconvenience of erroneously recognizing information that should not exist can be eliminated. Further, it becomes possible to surely include the information necessary for designating all the recording information of the medium in the voice recognition dictionary, and it is possible to eliminate the inconvenience that the necessary information cannot be recognized.

[0011]

BEST MODE FOR CARRYING OUT THE INVENTION An embodiment of the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram showing the arrangement of the voice recognition dictionary creating apparatus according to the present embodiment. In the embodiment shown in FIG. 1, the voice recognition dictionary creating apparatus of the present invention is incorporated as a part of a voice recognition unit.

In FIG. 1, 10 is a voice recognition unit,
Reference numeral 20 denotes an audio unit, which are connected via a bus 30 to exchange data with each other. The audio unit 20 performs a process of reproducing music recorded on a CD / MD / cassette tape or the like, or reproducing a received radio broadcast. The voice recognition unit 10 is for performing control of the audio unit 20 (for example, an operation of designating music to be reproduced) by voice recognition, and generates a control command corresponding to an input voice.

Although only the audio unit 20 is shown as an electronic device to be controlled by voice recognition here, an air conditioner, a navigation device and other electronic devices can also be used. In that case, those electronic devices are also connected to the bus 30 and exchange data with the voice recognition unit 10.

Next, the internal structure of the voice recognition unit 10 will be described. Reference numeral 11 denotes a voice input unit, which inputs a word representing a desired control content uttered by the user for controlling the audio unit 20 through a microphone. 12
Is a speech recognition dictionary, which is a database of recognition target vocabulary created by collecting recognition data for each word to be recognized.

In the present embodiment, an example in which a track number of a CD is recognized by voice and a desired music is selected will be described.
In this case, the voice recognition dictionary 12 includes audio commands such as "Seedee <track number>"
The association with the command ID corresponding to each audio command is stored. However, this stored content is not fixed in advance as in the past, but is dynamically created according to the recorded content for each CD as described later.

Reference numeral 13 is a matching unit, which is a voice input unit 1.
The user's voice input by 1 is compared with the recognition data stored in the voice recognition dictionary 12 to recognize which word is uttered. Then, the recognition result is output to the command output unit 14 in the form of a unique command ID assigned to each word. Command output unit 14
Is the command ID supplied from the matching unit 13,
It is converted into a corresponding control command and is sent to the audio unit 20 to be controlled via the bus 30.
This realizes the control commanded by the user by voice.

Reference numeral 15 is an attribute information input section, which obtains attribute information relating to the recorded contents of the CD from the audio unit 20 via the bus 30. The attribute information to be acquired is, for example, the number of tracks representing the number of songs recorded on the CD.
Specifically, when a CD is inserted into the audio unit 20, the attribute information input unit 15 acquires the number of tracks on that CD. The number of tracks is unique to each CD, and in most cases, the number of tracks is about 20 at most.

Reference numeral 16 is a dictionary creating unit, which is the attribute information (the number of CD tracks) acquired by the attribute information input unit 15.
Based on the above, the voice recognition dictionary 12 is created within the range of the number of commands corresponding to the number of tracks. FIG. 2 is a diagram showing an example of creating the voice recognition dictionary 12. As shown in FIG. 2A, in the voice recognition dictionary 12, the track number and the corresponding command ID are empty in the initial state. If a CD containing 15 pieces of music is inserted in the audio unit 20 in this state, the attribute information input unit 15 acquires the number of tracks of “15” from the audio unit 20. In response to this, the dictionary creation unit 16 is
As shown in (b), the voice recognition dictionary 12 that includes only track numbers 1 to 15 and the corresponding command IDs is created.

FIG. 3 is a flow chart showing the operation of the speech recognition dictionary creating apparatus according to the present embodiment having the above-mentioned structure. In FIG. 3, when the power is turned on, the dictionary creation unit 16 of the voice recognition unit 10 causes the voice recognition dictionary 12
Is initialized to the state of FIG. 2 (a) (step S1). Then, the audio unit 20 determines whether or not a medium such as a CD has been inserted (step S2). When the CD is inserted, the attribute information of the inserted CD is transmitted from the audio unit 20 to the voice recognition unit 10 (step S3).

The attribute information transmitted from the audio unit 20 is acquired by the voice recognition unit 10 via the attribute information input section 15. Based on the attribute information acquired by the attribute information input unit 15, the dictionary creating unit 16
In the range of the number of commands according to the attribute information, FIG.
Create a voice recognition dictionary 12 as shown in (step S
4). The voice recognition dictionary 12 created in this way is
It is used for the voice recognition processing by the matching unit 13.

After the voice recognition dictionary 12 is created in step S4, or when it is not detected that the CD is inserted in the audio unit 20 in step S2, the audio unit 20 determines whether the CD is removed. It is determined (step S5). If the CD is removed, the process returns to step S1 and the voice recognition dictionary 12 is displayed as shown in FIG.
Initialize to the state of (a). On the other hand, if the CD is not removed, the process returns to step S2 and the loop process of steps S2 to S5 is continuously performed.

As described in detail above, according to this embodiment, when a music CD is inserted into the audio unit 20, the number of tracks of the CD is acquired from the audio unit 20 and the track number of the CD is held. In this range, the voice recognition dictionary 12 is dynamically created.
As a result, it becomes possible to remove the track numbers that do not exist in the inserted CD from the voice recognition dictionary 12 in advance, including the track numbers that may be erroneously recognized as the actually existing track numbers. , It is possible to eliminate the inconvenience of erroneously recognizing a track number which should not exist on the CD.

For example, when a desired music piece is selected by voice recognition for a CD having track numbers 1 to 15, "fiftee" is used for the user to specify the track number.
Suppose you pronounced "n". In this case, since the track number of "fifty" whose pronunciation is similar to that of "fifteen" does not exist in the voice recognition dictionary 12, it is possible to prevent the matching unit 13 from erroneously recognizing a track number that does not actually exist. be able to. Therefore, a troublesome work such as redoing voice input is unnecessary, and the usability of voice recognition is greatly improved.

Further, according to the above-mentioned embodiment, the voice recognition dictionary 1 can reliably execute the command necessary for designating the music recorded on the CD inserted in the audio unit 20.
2 can be included. For example, the number of tracks is 15
When the CD of is inserted into the audio unit 20,
When the voice recognition dictionary 12 including the track numbers 1 to 15 is dynamically created and a CD having 20 tracks is inserted into the audio unit 20, the voice recognition dictionary 12 including the track numbers 1 to 20 is inserted. Is dynamically created. Therefore, unlike the conventional example in which a fixed voice recognition dictionary in which the number of commands is narrowed down is created in advance, it is possible to eliminate the inconvenience of being unable to recognize the track numbers of the music pieces existing on the inserted CD.

In the above embodiment, an example in which a CD is inserted into the audio unit 20 has been described, but the same applies to the case where another medium such as an MD, cassette tape or DVD is inserted. That is, the voice recognition unit 10 acquires the number of tracks of various media inserted in the audio unit 20, and dynamically creates the voice recognition dictionary 12 within the range of track numbers of the media.

In the above embodiment, an example in which the attribute information acquired from the audio unit 20 to the voice recognition unit 10 is the number of tracks on the medium has been described.
The attribute information to be acquired is not limited to this. For example, it may be the title of each song recorded on the medium. In this case, the dictionary creating unit 16 creates the voice recognition dictionary 12 within the range of the number of commands according to the number of acquired titles.

In the audio unit 20,
For example, MP3 (MPEG-Audio Layer3), AAC (Advance
d Audio Coding), AC-3 (Dolby Digital), ATR
AC3 (Adaptive Transform Acoustic Coding), Twin
When music compressed by VQ, WMA (Windows Media Audio: Windows is a registered trademark) or the like can be played, the file name of music data may be used as the attribute information.

Further, in the above embodiment, an example in which the track number of the medium is voice-recognized and desired music is selected has been described, but the recognition target is not limited to the track number. For example, the title of the music may be recognized by voice to select the music. In this case, the attribute information input unit 1
5 acquires the title of each song recorded on the medium from the audio unit 20. Further, the dictionary creating unit 16 creates the voice recognition dictionary 12 in which each of the acquired titles and the command ID corresponding thereto are associated with each other.

Generally, the title of a music piece varies from medium to medium. Therefore, if an attempt is made to create a voice recognition dictionary for voice recognition of a large number of media by song names, the number of commands (the number of song titles) required in the database of the voice recognition dictionary will be enormous. Also, many media are provided at any time, and it is necessary to successively update the voice recognition dictionary in order to perform voice recognition corresponding to the latest media. Therefore, it is not realistic to recognize the voice based on the title of the music.

On the other hand, according to this embodiment, according to the medium inserted in the audio unit 20, the voice recognition dictionary 12 using the title of the music recorded on the medium as a command is dynamically created. It Therefore, the number of commands required for the voice recognition dictionary 12 is only the number of music pieces recorded in the inserted medium, and it is not necessary to prepare a huge database. Further, it is not necessary to manage the update of the database, and voice recognition can be performed on any media based on the song name.

In the above embodiment, the voice recognition dictionary 1
In the initial state of 2, the track number and the command ID corresponding to it are left empty as shown in FIG.
The example of adding the track number and the like as needed when the medium is inserted has been described, but the present invention is not limited to this. For example, in the initial state of the voice recognition dictionary 12, track numbers 1 to 99 and command IDs corresponding thereto are prepared as shown in FIG. 4, and unnecessary track numbers and the like when a medium is inserted (for example, When a CD having 15 tracks is inserted, 16 to 99 parts) may be masked. However, even if it is said to be a mask, the processing by the matching unit 13 is 1 to 9
It is necessary to compare all 9 search spaces.
Therefore, the search space becomes smaller and the recognition speed can be made faster by adding only the necessary commands later as in the above embodiment.

Further, in the above-described embodiment, the voice recognition unit 10 and the audio unit 20 (further, various electronic devices such as an air conditioner and a navigation device, not shown) are connected to each other by the bus 30, and the voice of the present invention is applied to the system. Although the example of applying the recognition dictionary creating device has been described, the invention is not limited to this. For example, the voice recognition dictionary creating device of the present invention may be applied to a single audio device having a voice recognition function. Moreover, the applied system is not limited to the vehicle-mounted type.

In addition, the above embodiment is merely an example of the embodiment for carrying out the present invention, and the technical scope of the present invention should not be limitedly interpreted thereby. That is, the present invention can be implemented in various forms without departing from the spirit or the main features thereof.

[0034]

As described above, according to the present invention,
It does not exist because the attribute information related to the recorded information of the medium is acquired from the electronic device that is controlled by the voice recognition and the voice recognition dictionary is created within the range of the number of commands according to the acquired attribute information. It is possible to improve the usability of the voice recognition without erroneously recognizing the supposed information or not being able to recognize the supposed information.

[Brief description of drawings]

FIG. 1 is a block diagram showing a configuration of a voice recognition dictionary creating device according to an embodiment.

FIG. 2 is a diagram showing an example of creating a voice recognition dictionary.

FIG. 3 is a flowchart showing an operation of the voice recognition dictionary creating device according to the present embodiment.

FIG. 4 is a diagram showing an example of a conventional voice recognition dictionary.

[Explanation of symbols]

10 Speech recognition unit 11 Voice input section 12 Speech recognition dictionary 13 Matching part 14 Command output section 15 Attribute information input section 16 Dictionary creation department 20 audio units 30 bus

Claims (7)

[Claims] 1. A device for creating a voice recognition dictionary used when designating any one of record information of a medium used in an electronic device by voice recognition, the device relating to record information of the medium from the electronic device. Attribute information acquisition means for acquiring attribute information, and dictionary creation means for creating a voice recognition dictionary within the range of the number of commands according to the attribute information based on the attribute information acquired by the attribute information acquisition means A voice recognition dictionary creating device characterized by the above. 2. The voice recognition dictionary creating apparatus according to claim 1, wherein the attribute information is the number of pieces of record information recorded in the medium. 3. The voice recognition dictionary creating apparatus according to claim 1, wherein the attribute information is a title of each piece of recorded information recorded on the medium. 4. The voice recognition dictionary creating apparatus according to claim 1, wherein the attribute information is a file name of each of the record information recorded on the medium. 5. The dictionary creating means creates a voice recognition dictionary using the title as a command, based on the title of each record information acquired by the attribute information acquiring means. The described voice recognition dictionary creation device. 6. The dictionary creating means creates a voice recognition dictionary using the file name as a command, based on the file name of each record information acquired by the attribute information acquiring means. 4. The voice recognition dictionary creation device described in 4. 7. A method for creating a voice recognition dictionary used when designating any one of recorded information on a medium used in an electronic device by voice recognition, the method comprising: A voice recognition dictionary creating method, characterized in that attribute information is acquired and a voice recognition dictionary is created within a range of the number of commands according to the acquired attribute information.