JPH0926799A - Speech recognition device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a speech recognition device which can efficiently use the storage capacity of a dictionary and is good in operability. SOLUTION: When one operation is specified for the first time on a touch panel, etc., the standard pattern of the words indicating a specified operation is generated based on phoneme patterns representing features of respective phonemes stored in a dictionary part 167 and stored in a specific dictionary 163b in a dictionary buffer 163. A feature extraction part 162 extracts features from an input speech from a microphone 24 and generates a word pattern, and a pattern matching part 165 compares it with standard patterns stored in a RAM 164 and the dictionary buffer 163 to recognize the speech. Further, the word pattern generated by the feature extraction part 162 is stored as a standard pattern for a specific speaker in the RAM 164 and the specific dictionary 163b. Since a user registers standard patterns of only necessary words, an efficient dictionary is generated and an object of speech recognition can easily be understood.

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 more particularly, to a voice recognition device for recognizing a voice uttered for a specific phrase.

[0002]

2. Description of the Related Art Speech recognition devices for recognizing speech spoken by humans as words have been put to practical use in various fields. This voice recognition device has been put to practical use as an input device for giving a voice instruction from a place where an instruction corresponding to various devices in a factory has been released. It is considered to be used as a voice input device when inputting. In such a voice recognition device, in general, in order to identify the input voice, by analyzing the frequency distribution of the voice to be recognized in advance, for example, the spectrum or the time series information of the fundamental frequency is extracted as a feature. , A voice recognition dictionary for storing the pattern in association with each word is provided. As a dictionary stored in this voice recognition dictionary, a pattern obtained by performing averaging processing on the frequency distribution of the voice for an unspecified large number of speakers is generally used as a pattern (the following is referred to as a standard pattern). There are an unspecified speaker dictionary in which patterns are registered with words, and a specified speaker dictionary in which patterns of input speech are registered together with words for a specific speaker.

When a voice to be recognized is input, the frequency pattern of the input voice and the pattern of each word stored in both dictionaries are compared and collated by pattern matching, and the similarity to each word is calculated. Next, the word with the highest calculated similarity (the word with the closest pattern) is recognized as the input voice, and that word is output. That is, the input voice is determined by checking which word pattern most closely matches the pattern of the frequency distribution of the input word.

[0004]

However, the current speech recognition apparatus has a limit in the storage capacity of standard patterns in the unspecified dictionary and the specified dictionary, and when the number of vocabularies increases, many standard patterns are accumulated in both dictionaries. could not. As described above, since the storage capacity of the standard pattern is limited, only the standard pattern for the preselected word is registered in the unspecified dictionary. However, the selected registered word is not always necessary for the user, and when it is an unnecessary word, the storage capacity used is wasted. Also, what kind of word in advance in the dictionary has been registered, for the user, did not know unless you see the words that have been registered in the use manual or the like. For this reason,
The conventional voice recognition device has a problem in terms of user operability.

Therefore, the present invention has been made to solve the above problems, and an object of the present invention is to provide a voice recognition device which can efficiently use the storage capacity of a dictionary and has good operability.

[0006]

According to a first aspect of the present invention, there is provided a selecting means for selecting a predetermined operation, and a phoneme pattern representing features of each phoneme that constitutes a word representing an operation selectable by the selecting means. A phoneme dictionary in which is stored in advance,
Dictionary storage means for storing a standard pattern, standard pattern creating means for creating a standard pattern of a word representing the operation selected by the selecting means from the phoneme patterns stored in the phoneme dictionary, and the standard pattern creating means The unspecified dictionary storage means for storing the standard pattern created in 1. in the dictionary storage means, the voice input means for inputting voice, and the feature of the voice input by this voice input means are extracted to obtain a corresponding word pattern. A word pattern creating means for creating, and a recognizing means for recognizing the voice input by the voice input means based on the similarity between the word pattern created by the word pattern creating means and the standard pattern stored in the dictionary storage means. The above-mentioned object is achieved by providing a voice recognition device with an output means for outputting the recognition result by this recognition means.
According to a second aspect of the present invention, in the voice recognition apparatus according to the first aspect, a specific dictionary that stores the word pattern created by the word pattern creating means in the dictionary storage means as a standard pattern for a specific speaker. Storage means is further provided. In the invention according to claim 3, in the voice recognition device according to claim 1 or 2, when the number of standard patterns stored in the dictionary storage means reaches a predetermined number,
The unspecified dictionary storage means and the specified dictionary storage means delete the standard pattern with the least recognition frequency by the recognition means. According to the invention described in claim 4, the voice recognition device according to claim 1, claim 2, or claim 3 is used as an input unit of the navigation device.

[0007]

In the speech recognition apparatus according to the first aspect of the present invention, the phoneme pattern representing the characteristics of each phoneme forming the word representing the operation selectable by the selecting means is stored in advance in the phoneme dictionary. Then, a standard pattern of the word representing the operation selected by the selection means is created from the phoneme pattern and stored in the dictionary storage means. On the other hand, the feature of the voice input by the voice input means is extracted to create a corresponding word pattern, and the input voice is recognized based on the similarity with the standard pattern stored in the dictionary storage means, and the recognition result is obtained. Output. In the voice recognition device according to the second aspect, the word pattern created from the input voice is stored in the dictionary storage means as a standard pattern for a specific speaker. In the voice recognition device according to the third aspect, when the number of standard patterns stored in the dictionary storage means reaches a predetermined number, the standard pattern having the lowest recognition frequency is deleted. In the voice recognition device according to the fourth aspect, it is used as an input unit of the navigation device.

[0008]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the voice recognition apparatus of the present invention will be described in detail below with reference to FIGS. Outline of Example In the speech recognition apparatus of this example, when a certain operation is designated for the first time on a touch panel or the like, a standard pattern of a word indicating the designated operation is created based on a phoneme pattern representing characteristics of each phoneme. Then, it is stored in the dictionary for unspecified speakers. When a voice is input, the features of the input voice are extracted to create a word pattern, and the voice is recognized based on the similarity to the standard pattern stored in the dictionary. The word pattern created at the time of this voice recognition is stored in the dictionary as a standard pattern for a specific speaker. As described above, in the present embodiment, the standard pattern is created only for the necessary word once specified from the touch panel or the like, and it becomes possible to input it by voice from the next time. Since the word that is the target of voice recognition is the word that is once designated and input by the user, the user does not need to check the recognition target again.

Details of Embodiment FIG. 1 shows a system configuration when a voice recognition device according to an embodiment of the present invention is applied to a navigation device. This navigation device includes a calculation unit 10. The calculation unit 10 includes a display unit 11 including a display 11a that functions as a touch panel and operating switches 11b provided around the display 11a, and a touch panel and switches 11 of the display unit 11.
A switch input management unit 12 that manages the input from b is connected.

The switch 11b includes various switches such as a switch for designating a navigation menu screen, a switch for adjusting an air conditioner, and a switch for operating an audio. When these switches are pressed, the corresponding menu screen is displayed on the display 11a. The screen displayed on the touch panel 11a has a hierarchical structure, and the menu screen is at the top layer. And on the menu screen for navigation,
For example, a designation key for designating a destination, searching for a place name, etc. is displayed, and when the destination setting key is designated, a designation key for designating a ski resort, a golf course, etc. is displayed as a screen lower than that. The screen of the lowest layer is sequentially displayed by the designation of each designation key. In this embodiment, the contents of the various designation keys displayed on the screen of the touch panel 11a and the various switches of the switch 11b are the targets of voice recognition. That is, every time these various designated keys and switches are first pressed, an unspecified speaker dictionary for voice recognition is created,
After that, the input voice is recognized based on the created dictionary, and the specific speaker dictionary based on the input voice is created.

Further, the arithmetic unit 10 includes a current position measuring unit 1
3, a speed sensor 14, a map information storage unit 15, a voice recognition unit 16 in this embodiment, and a voice output unit 17 are connected. The current position measuring unit 13 detects the current position where the vehicle is currently running or stopped by detecting coordinate data based on latitude and longitude. This current position measurement unit 1
3 is a GP that measures the position of vehicles using artificial satellites
An S (Global Position System) receiver 21, a beacon receiving device 20 that receives position information from beacons placed on the road, an azimuth sensor 22, and a distance sensor 23 are connected, and the current position measuring unit 13 is connected to them. The current position of the vehicle is measured using the information of.

The azimuth sensor 22 is, for example, a geomagnetic sensor that detects the geomagnetism to determine the azimuth of the vehicle, a gas rate gyro or an optical fiber gyro that detects the rotational angular velocity of the vehicle and integrates the angular velocity to determine the azimuth of the vehicle. gyro,
A wheel sensor or the like is used in which left and right wheel sensors are disposed, and a displacement of the azimuth is calculated by detecting turning of the vehicle based on an output pulse difference (difference in moving distance). As the distance sensor 23, for example, various methods such as a method of detecting and counting the number of rotations of a wheel, or a method of detecting acceleration and integrating twice are used. Note that the GPS receiver 21 and the beacon receiving device 20 can perform position measurement independently,
In a place where reception by the PS receiver 21 or the beacon receiving device 20 is not possible, the current position is detected by dead reckoning navigation using both the direction sensor 22 and the distance sensor 23.

The map information storage unit 15 is, for example, a CDROM.
And the like. The map information storage unit 15 stores various data necessary for route search and route guidance, such as road data necessary for route search to the destination and map data for displaying the searched route on the display 11a. Have been. The voice recognition unit 16 is connected to a microphone 24 into which a human voice or a dial tone corresponding to a telephone number is input. The audio output unit 17 includes an audio output IC 26 for outputting audio as an electric signal, a D / A converter 27 for digital-to-analog conversion of the output of the audio output IC 26, and an amplifier 28 for amplifying the converted analog signal. It has. A speaker 29 is connected to an output terminal of the amplifier 28.

The arithmetic unit 10 includes a CPU (central processing unit),
A ROM (Read Only Memory), a RAM (Random Access Memory), etc. are provided, and the CPU executes the program stored in the ROM using the RAM as a working area to realize each of the above configurations. ing. That is, the calculation unit 10 manages the map data reading unit 31, the map drawing unit 32, and the map data reading unit 31 and the map drawing unit 32 which are connected to the speed sensor 14 and the map information storage unit 15. And a screen management unit 34 connected to the map drawing unit 32 and the display unit 11.
An input management unit 35 connected to the switch input management unit 12 and the voice recognition unit 16, a voice output management unit 36 connected to the voice output IC 26 of the voice output unit 17, a map management unit 33, and a screen. Management unit 34, input management unit 3
5, an overall management unit 37 that manages the audio output management unit 36. The input management unit 35 includes various designation keys displayed on the touch panel of the display 11a and the switch 1
It also manages whether 1b is the first one selected. When a switch or the like is selected for the first time, the voice recognition unit 16 is supplied with an input signal such as a touch panel corresponding to the switch or the like to the input management unit 35.

FIG. 2 is a block diagram showing the configuration of the voice recognition unit 16 in FIG. As shown in this figure, the voice recognition unit 16 includes an A / D conversion unit 161 for converting a voice signal input from the microphone 24 into a digital signal, and the D / D conversion unit 161.
A / A converter 161 extracts a feature of an input voice from an output signal and creates a corresponding word pattern, and a feature extraction unit 162, and a dictionary buffer 163 as a dictionary that stores a standard pattern of a predetermined voice. And RAM16
4 is provided. The dictionary buffer 163 stores an unspecified dictionary 163a in which a general standard pattern for recognizing an unspecified speaker is stored, and a word pattern created from the sound of the specified speaker input from the microphone 24 as a standard pattern. It has a specific dictionary 163b. Here, the standard pattern refers to an audio signal for each predetermined unit,
It is time series information of spectrum and fundamental frequency. As the voice signal for each predetermined unit, a voice signal for each unit such as syllable, word, phoneme, semi-syllable, inter-word, inter-phoneme, inter-syllable, etc. is used. Further, the feature extraction unit 162 is adapted to extract this word pattern by a multi-channel bandpass filter, a linear prediction analysis, or the like.

Further, the voice recognition unit 16 uses the dictionary buffer 1
63, the pattern matching unit 165 that compares the standard pattern stored in the RAM 164 with the word pattern extracted by the feature extracting unit 162, and the voice input from the microphone 24 based on the comparison result of the pattern matching unit 165. The content is recognized, a voice input signal to the calculation unit 10 is generated according to the recognition content, and the calculation unit 10 is generated.
Of the recognition result processing unit 166 for outputting to the input management unit 35.

Further, the voice recognition unit 16 has a dictionary unit 167.
And a ROM 168 and a dictionary management unit 169. The dictionary unit 167 is composed of a ROM and stores phoneme patterns representing the characteristics of each phoneme. This phoneme pattern is used to create an unspecified dictionary by the dictionary management unit 169, and is stored for each phoneme by averaging voicing sounds by a plurality of announcers and the like.
Although the ROM of the dictionary unit 167 can use various storage devices, a CD (compact disc) RO
You can use M.

The input management unit 35 supplies the ROM 168 with input signals such as a designated key displayed on the touch panel of the display 11a and a touch panel corresponding to the switch of the switch 11b, which are pressed by the user. . The ROM 168 is adapted to convert into a phoneme designating signal indicating a word corresponding to an input signal such as a touch panel and supply the phoneme designating signal to the dictionary management unit 169, and is provided with a conversion table therefor. For example, when the "destination setting" designation key is pressed from the touch panel of the display 11a and a corresponding touch panel input signal is supplied from the input management unit 35, the ROM 168 stores the word "corresponding to the touch panel input signal". The dictionary management unit 169
To supply.

The dictionary management unit 169 reads out each phoneme pattern corresponding to the phoneme designating signal supplied from the ROM 168 from the dictionary unit 167, synthesizes a dictionary for unspecified speaker recognition, and stores this in the dictionary buffer 163. Specific dictionary 163a
It is designed to be stored in. When the recognition of the voice input from the microphone 24 is successful, the dictionary management unit 169 is supplied with the voice input signal from the recognition result processing unit 166. When this voice input signal is supplied, the dictionary management unit 169 stores the word pattern extracted by the feature extraction unit 162 for the recognized voice as a standard pattern in the specific dictionary 163b and the RAM 164. Here, the RAM 164 is provided with an area for storing as many words as the switches that can be pressed on the display unit 11 and designated keys (= the number of voice input signals), and one standard pattern is provided in each area. Is stored. Therefore, every time the voice input from the microphone 24 is recognized, the storage area corresponding to the input voice is updated with the new standard pattern. The phoneme pattern of the input voice is also updated in the RAM 164 as a phoneme dictionary. That is, each time the voice input from the microphone 24 is recognized, a phoneme pattern for each phoneme is created from the word pattern extracted by the feature extraction unit 162 for the input voice, and the RAM 164 is updated.

The specific dictionary 163b also has an area for storing as many words as the number of voice input signals, and a plurality of standard patterns are stored in each area. Therefore, every time the voice is recognized, a new standard pattern is sequentially stored in the storage area corresponding to the input voice. After the predetermined number of standard patterns are stored in the storage area corresponding to the voice input signal, the update is performed between the oldest standard pattern and the latest standard pattern.

Next, the operation of the embodiment thus constructed will be described. FIG. 3 shows a usage state of the navigation device by the user. The navigation device of the present embodiment has the RA in the initial state of the system.
M164 and the unspecified dictionary 163a in the dictionary buffer 163,
The standard pattern is not stored in the specific dictionary 163b. Therefore, voice recognition cannot be performed in the initial state of the system, and as shown in FIGS. 3A to 3D, the switch 11b and the display 11a of the display device 11 are sequentially displayed.
Select a predetermined process from the touch panel. Here, a case where the destinations are sequentially input from the navigation menu screen (a) using the touch panel will be described. First, when the user selects "destination setting" on the menu screen (a), the destination setting screen (b) of the lower layer
Is displayed. Similarly, by selecting "Golf course" on the destination setting screen (b), and further sequentially selecting "Chiba prefecture", "Narashino CC", etc.
Narashino Country Club is set as the destination by 7.

The display 11a of the display unit 11
Each time the switch of the touch panel or the switch 11b is selected for the first time, a standard pattern of words corresponding to the switch is created by the dictionary management unit 169, and the dictionary buffer 16
3 in the unspecified dictionary 163. That is, FIG.
When the selection is performed in the order of (a) to (d), the standard patterns of the words "destination setting", "golf course", "Chiba prefecture", and "Narashino country club" corresponding to the selection are sequentially stored in the specific dictionary 163a. To be done.

In this way, the destination is set by the operation of the display unit 11, and at the same time, the unspecified dictionary of the words indicated by the corresponding designated key or switch is created, and thereafter, the voice input is performed. Will be possible. That is, on the navigation menu screen shown in FIG. 3A, when the user utters "destination setting" and inputs it into the microphone 24 as shown in (e), the voice recognition unit 16 recognizes it and displays it on the display 11a. Displays a destination setting screen (b). Further, when the user utters "golf course" (FIG. 3 (f)), a golf course screen is displayed. Similarly,
When the user utters “Chiba Prefecture” and “Narashino Country Club” in that order, the voice is recognized, and finally Chirashino CC is set as the destination.

Here, with respect to the voice recognized and uttered by the user, the word pattern of the voice is stored in the specific dictionary 136b and the RAM 164 as a standard pattern for a specific speaker. In FIG. 3G, if the user does not say "Chiba Prefecture" but "Hokkaido", the unspecified dictionary of the word Hokkaido has not been created yet.
I can't recognize. In order to enable recognition of the voice "Hokkaido", it is necessary to once select the "Hokkaido" key displayed on the golf course screen of FIG. 3C or another screen.

Next, details of the operation will be described. FIG.
FIG. 6 is a flowchart showing an operation of dictionary creation and voice recognition by the created dictionary. First, the control unit 10
It is determined whether or not data is input to the input management unit 35,
When there is an input (step 31; Y), it is determined whether the input data is an input from the touch panel of the display 11 or the switch 11b (step 32).

In the case of input from the touch panel or the like (step 32; Y), the input management unit 35 determines whether or not the input key has been used for the first time, and thereby the word representing the selected key is displayed. It is judged whether or not the standard pattern of is in the unspecified dictionary (step 33). If there is no unspecified dictionary to be selected (step 33; Y), the voice recognition unit 16
Creates an unspecified dictionary to be selected (step 3
4). That is, the ROM 168 of the voice recognition unit 16 converts the input signal such as the touch panel supplied from the input management unit 35 into a phoneme designation signal indicating the corresponding word and supplies the phoneme designation signal to the dictionary management unit 169. The dictionary management unit 169 reads out the phoneme pattern of each phoneme from the dictionary unit 167 according to the supplied phoneme designating signal, creates the unspecified speaker dictionary for the word representing the key input in step 32, and creates the unspecified dictionary 163. To store.

After the unspecified dictionary is created in step 34, or when there is an unspecified dictionary to be selected (step 33; N), the input management unit 35 inputs the input from the touch panel or the like to the general management unit 37. Supply (Step 3)
5). The overall management unit 37 controls the screen management unit 34 in response to an input from the touch panel or the like, switches to a screen for selecting the next vocabulary to be recognized (step 36), and returns to the main routine.

On the other hand, in step 32, when the input is not a touch panel input but voice input (; Y),
The voice recognition unit 16 inputs the voice uttered by the user from the microphone 24 (step 37), and performs voice recognition on the input voice (step 38). That is, the voice recognition unit 16 converts the input voice data into the A / D conversion unit 1
The data is converted into digital data at 61, the word pattern of the input voice is extracted at the feature extraction unit 162, and the word pattern is supplied to the pattern matching unit 165 and the dictionary management unit 169.

The pattern matching unit 165 first compares the word pattern of the input voice with the latest standard pattern for each word stored in the RAM 164.
The degree of similarity with each word is calculated, and the word with the highest degree of similarity is recognized as the input voice. If the similarity with any of the words is less than or equal to a predetermined threshold value, it cannot be recognized. Therefore, the similarity is calculated by comparing with the standard pattern of each word stored in the specific dictionary 163b. Recognize that the word that is larger than the threshold and has the highest degree of similarity is speech.
If all the similarities to the standard pattern of the specific dictionary 163b are also less than or equal to the threshold value, further comparison with all the standard patterns for the unspecified speaker stored in the specific dictionary 163a is performed, and the similarity is larger than the threshold value. A word having a high degree of similarity is recognized as a voice, and if both are below a threshold, it is determined that the input voice cannot be recognized. When the recognition result by the pattern matching unit 165 is obtained, the recognition result processing unit 166 generates a voice input signal according to the recognition content and supplies it to the input management unit 35 of the calculation unit 10.

The overall management unit 37 of the control unit 10 is
When the recognition result supplied to the input management unit 35 is unrecognizable (step 39; N), the screen management unit 34 is controlled to display on the display 11a that there is no dictionary corresponding to the input voice. Display (step 40),
Return to the main routine. On the other hand, when the similarity with the standard pattern is larger than the threshold and the recognition of the input voice is successful (step 39; Y), the specific dictionary is updated (step 41). That is, the recognition result processing unit 166 outputs the voice input signal corresponding to the recognized word to the dictionary management unit 16
9 is also supplied. When the voice input signal is supplied, the dictionary management unit 169 updates the storage area of the RAM 164 corresponding to the voice input signal with the word pattern supplied from the feature extraction unit 162. Further, the dictionary management unit 169
The word pattern supplied from the feature extraction unit 162 is stored in the corresponding storage area of the specific dictionary 163b, or the oldest standard pattern stored in the storage area is rewritten to the supplied word pattern.

When the updating of the specific dictionary is completed in step 41, the process proceeds to step 35, and the overall management unit 37
Supplies the voice input signal of the voice input from the microphone 24 and recognized by the voice recognition unit 16 to the overall management unit 37.
Then, the overall management unit 37 controls the screen management unit 34 according to the voice input signal, switches to the screen for selecting the next vocabulary to be recognized (step 36), and returns to the main routine.

In the embodiment described above, the RAM 1
64, the unspecified dictionary 163a, and the specified dictionary 163 have an area for storing words corresponding to the number of switches and designated keys that can be pressed on the display unit 11, but a smaller number, for example, half or 1 / It may be a storage area of 3. In this case, the dictionary management unit 169 counts the usage frequency of the words indicated by the standard pattern stored in each area in response to the area shortage, and determines the storage area with the least usage frequency. Clear and set aside for entering new word patterns or standard patterns. As a result, it is possible to reduce the storage capacity while suppressing a decrease in the recognition rate. Also, pattern matching 1
Since the number of dictionaries compared at 65 is reduced, the recognition speed can also be improved. Further, even when the number of recognized vocabularies and the storage capacity of the dictionary buffer and the like are limited, an efficient system can be realized by reconstructing the dictionary by using the dictionary.

Although the phoneme designating signal output from the ROM 168 is output, data indicating a text character string may be used as the phoneme designating signal.

Further, the screen pointer for specifying the screen displayed by the input of the voice may be stored in association with each of the standard patterns for the recognized voice so that the screen can be called directly. . For example, in the case where FIG. 3B is displayed as a screen corresponding to the case where the voice "destination setting" is recognized, FIG. 3 is displayed in association with the standard pattern of the voice "destination setting".
A screen pointer that specifies (b) is also stored. Similarly, the screen pointer of FIG. 3C is stored in association with the standard pattern of the voice “Golf course”, and the screen pointer of FIG. 3D is also stored in association with the standard pattern of the voice “Chiba Prefecture”. It After that, even if voice input and recognition are not performed in order of destination setting, golf course, Chiba prefecture, and Narashino CC, by voice inputting "golf course", the screen pointer shown in FIG.
The screen of (c) is displayed. Then, by voice-inputting “Narashino CC”, Narashino CC is set as the destination, and the input process becomes easy. Further, instead of storing the standard pattern of voice together with the screen pointer for specifying the screen, a series of input processing procedures and input history may be stored together with the standard pattern of voice. As a result, in the case of the example in FIG. 3, the destination setting is completed by directly outputting the voice "Narashino CC", and the input processing can be performed more easily and quickly.

[0035]

According to the voice recognition apparatus of the present invention, the storage capacity of the dictionary can be used efficiently and the operability can be improved.

[Brief description of drawings]

FIG. 1 is a system configuration diagram when a voice recognition device according to an embodiment of the present invention is applied to a navigation device.

FIG. 2 is a block diagram showing a configuration of a voice recognition unit of the above.

FIG. 3 is an explanatory diagram showing a usage state of a navigation device by a user.

FIG. 4 is a flowchart showing an operation of creating a dictionary and voice recognition by the created dictionary.

[Explanation of symbols]

 10 arithmetic unit 11 display unit 11a display 13 present 1 measurement unit 15 map information storage unit 16 voice recognition unit 161 A / D conversion unit 162 feature extraction unit 163 dictionary buffer 164 RAM 165 pattern matching unit 166 recognition result processing unit 167 dictionary unit 168 ROM 169 Dictionary management unit 17 Voice output unit 24 Microphone 33 Map management unit 34 Screen management unit 35 Input management unit 37 Overall management unit

─────────────────────────────────────────────────── ─── Continuation of the front page (51) Int.Cl. ⁶ Identification code Office reference number FI Technical display location G09B 29/10 G09B 29/10 A // G01C 21/00 G01C 21/00 H

Claims (4)

[Claims] 1. A selection unit for selecting a predetermined operation, a phoneme dictionary in which phoneme patterns representing features of each phoneme that constitutes a word representing an operation selectable by the selection unit are stored in advance, and a standard pattern is provided. Dictionary storing means to be stored, standard pattern creating means for creating a standard pattern of a word representing the operation selected by the selecting means from the phoneme patterns stored in the phoneme dictionary, and the standard pattern creating means. An unspecified dictionary storage means for storing the standard pattern in the dictionary storage means, a voice input means for inputting a voice, and a word for creating a corresponding word pattern by extracting features of the voice input by the voice input means. From the pattern creating means and the degree of similarity between the word pattern created by the word pattern creating means and the standard pattern stored in the dictionary storage means, Serial speech recognition means for recognizing speech inputted by the input means, the speech recognition apparatus characterized by comprising output means for outputting a recognition result by the recognizing means. 2. The specific dictionary storage means for storing the word pattern created by the word pattern creation means in the dictionary storage means as a standard pattern for a specific speaker, according to claim 1. Voice recognition device. 3. When the number of standard patterns stored in the dictionary storage means reaches a predetermined number, the unspecified dictionary storage means and the specified dictionary storage means delete the standard pattern with the least recognition frequency by the recognition means. The voice recognition device according to claim 1 or 2, wherein 4. The voice recognition device according to claim 1, wherein the voice recognition device is used as an input means of a navigation device.