JP4498906B2 - Voice recognition device - Google Patents

Description

  The present invention relates to a voice recognition device used for operating a navigation system with voice, for example, and more particularly to a technique for improving a voice recognition rate.

  2. Description of the Related Art Conventionally, a voice recognition device that recognizes voice spoken by a user and outputs a recognition result is known. In a normal voice recognition device, a recognition dictionary used in voice recognition processing is prepared, and a vocabulary to be recognized is stored inside the recognition dictionary. In the speech recognition process, the vocabulary spoken by the user is analyzed, and compared with the vocabulary stored in the recognition dictionary, the recognition result is presented.

  As a technique related to speech recognition, Patent Document 1 discloses a speech input identification method in which data to be input is classified into a plurality of groups in advance and stored in a dictionary. In this speech input identification method, characters corresponding to speech are classified into three groups, a clear sound group, a muddy sound group, and a semi-voiced sound group, stored in a dictionary, and recognition processing is performed using the dictionary selected by the switch. By doing so, it is possible to input a voice of any one specified character.

  Further, Patent Document 2 discloses an electronic dictionary device that enables dictionary lookup in real time using an operation method of dictionary lookup without having an alphabet button on a portable electronic dictionary. In this electronic dictionary device, one voice recognition button is used for a plurality of operation methods, and voice recognition start / stop and candidate presentation operations can be performed with a single button without using a plurality of switches. It is. When the voice recognition button is pressed for 1 second or more, voice input is performed. When the voice recognition button is pressed once within 0.5 seconds, the next candidate of the recognition result is presented at the time of erroneous recognition. The voice recognition button is pressed twice within 0.5 seconds. Voice recognition is interrupted by pressing.

JP-A-55-2040 JP-A-10-171492

  By the way, in the conventional speech recognition apparatus described above, the vocabulary that can be recognized by the speech recognition processing increases by performing speech recognition using a large number of recognition dictionaries, and it becomes possible to accept many recognition vocabularies at a time. . However, there is a problem that misrecognition occurs more as the vocabulary to be recognized increases.

  In the technique disclosed in Patent Document 1 described above, since characters can be input only one character at a time, it is not possible to input continuous words at a time. When inputting a plurality of characters, the number of characters in the input word It only takes time to repeat the operation. Further, in order to switch dictionaries, it is necessary to prepare as many switches as the number of dictionaries corresponding to each dictionary. In addition, the technology disclosed in Patent Document 2 simply enables recognition start, correction, and interruption by operating a voice recognition button, and improving the recognition rate by specifying or switching a dictionary to be used is not possible. Can not.

  The present invention has been made to solve the above-described problems, and an object of the present invention is to provide a speech recognition apparatus capable of recognizing a target vocabulary at a high recognition rate with a simple configuration.

The speech recognition apparatus according to the present invention includes a plurality of recognition dictionaries for recognizing speech, a recognition start button for starting speech recognition, and one of the plurality of recognition dictionaries according to the number of times the recognition start button is operated. A control means for enabling the voice recognition means for performing voice recognition using the recognition dictionary that is set effectively by the control means, and the control means has a recognition start button with one recognition dictionary selected. when kept pressed for a predetermined time or more, always enabled the selected recognition dictionary, the speech recognition means, Ru der to perform speech recognition using an effective and constantly enabled by recognition dictionary.

According to the present invention, after a dictionary to be used for speech recognition is selected by pressing the recognition start button, the selection is performed when the recognition start button is continuously pressed for a predetermined time or more with one recognition dictionary selected. and the recognition dictionary is enabled at all times, since it is configured so as to perform recognition processing, it is possible to reliably narrowing vocabulary by the selection of the recognition dictionary can recognize vocabulary of interest at a high recognition rate. In addition, the recognition dictionary used for speech recognition shares the recognition start button for starting speech recognition and is configured to be selected according to the number of times the recognition start button is pressed. Can be configured at low cost.

Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.
Embodiment 1 FIG.
FIG. 1 is a block diagram showing a configuration of a speech recognition apparatus according to Embodiment 1 of the present invention. This voice recognition device includes a voice recognition dictionary 1, a recognition dictionary management means 2, a voice recognition means 3, a control means 4, an output information control means 4, a manual input means 6, a key code discrimination means 7, a voice input means 8, and a screen output. A means 9, an audio output means 10, a remote controller (hereinafter abbreviated as “remote control”) 11, a microphone (hereinafter abbreviated as “microphone”) 12, a monitor 13 and a speaker 14. The remote control 11 is provided with a recognition start button 15 for starting voice recognition. When the recognition start button 15 is pressed, the recognition process is started, and the voice uttered toward the microphone 12 is recognized.

  The speech recognition dictionary 1 is composed of a plurality of recognition dictionaries. In the speech recognition apparatus according to the first embodiment, the speech recognition dictionary 1 is composed of four recognition dictionaries # 1 to # 4 classified according to vocabulary types. Specifically, the recognition dictionary # 1 includes a vocabulary (prefecture name, municipality name, etc.) used for an address, and the recognition dictionary # 2 includes a vocabulary (Tokyo Tower, etc.) used for a facility name. “Facility name dictionary”, recognition dictionary # 3 is “phone number dictionary” including vocabulary (numbers) used for telephone numbers, and recognition dictionary # 4 is “song name dictionary including vocabularies (song names) used for song names. ].

  The recognition dictionary management means 2 manages a plurality of recognition dictionaries # 1 to # 4 constituting the speech recognition dictionary 1. This recognition dictionary management means 2 holds a management table 21 as shown in FIG. The management table 21 stores a dictionary number (# 1 to # 4), a dictionary name (address, facility name, telephone number, song name) and state (invalid or valid). Then, in response to the dictionary switching request sent from the control means 4, one recognition dictionary to be recognized is selected, the state of the selected recognition dictionary is set to “valid”, and other recognition dictionaries are set to “ Set to Disable. Also, the recognition dictionary management means 2 reads the contents of the recognition dictionary set to “valid” and sends it to the voice recognition means 3 when the voice recognition is performed.

  The voice recognition means 3 executes voice recognition processing with reference to the voice recognition dictionary 1. The voice recognition unit 3 includes a voice analysis processing unit 31 and a matching processing unit 32. The voice analysis processing unit 31 analyzes the voice data sent from the voice input unit 8 when an instruction to start recognition is given from the control unit 4. The analysis result in the voice analysis processing unit 31 is sent to the matching processing unit 32. The matching processing unit 32 executes a matching process for comparing the analysis result sent from the voice analysis processing unit 31 with the vocabulary sent from the voice recognition dictionary 1 via the recognition dictionary management means 2. The recognition result obtained by the matching processing in the matching processing unit 32, specifically, the recognition vocabulary and its correct answer probability (hereinafter referred to as “score”) are sent to the control means 4.

  The control unit 4 controls the entire speech recognition apparatus by transmitting and receiving data to and from the recognition dictionary management unit 2, the speech recognition unit 3, the output information control unit 5, the manual input unit 6, and the key code determination unit 7. . Details of the control means 4 will be described later.

  The output information control means 5 controls screen display and audio output. The output information control unit 5 includes an output information table 51, a screen output generation unit 52, and an audio output generation unit 53. As shown in FIG. 3, the output information table 51 includes screen display characters and utterance example characters displayed on the monitor 13 and voice guidance characters output from the speaker 14 in correspondence with the number of information to be output. I remember it. The screen display characters and utterance examples stored in the output information table 51 are read by the screen output generation means 52, and the voice guidance characters are read by the voice output generation means 53.

  The screen output generation means 52 generates screen output data based on the screen display characters and utterance example characters read from the output information table 51 in accordance with screen display information (details will be described later) sent from the control means 4. Further, the screen output generation means 52 generates screen output data based on the recognized vocabulary sent from the control means 4. The screen output data generated by the screen output generation means 52 is sent to the screen output means 9. The voice output generation means 53 generates voice output data based on the voice guidance characters read from the output information table 51 in accordance with the screen display information sent from the control means 4. The voice output generation unit 53 generates voice output data based on the recognized vocabulary sent from the control unit 4. The audio output data generated by the audio output generation unit 53 is sent to the audio output unit 10.

  The manual input means 6 receives a key code sent from the remote controller 11 when a key on the remote controller 11 is pressed and generates a key event. The key event and key code generated by the manual input means 6 are sent to the control means 4. As a means for generating a key event and a key code, other input means such as a touch panel and a push button switch can be used instead of the remote controller 11.

  The key code discrimination means 7 discriminates the key code sent from the control means 4. For example, when the key code of the recognition start button 15 is sent from the control unit 4, the key code determination unit 7 determines that the key code corresponds to the recognition start button 15, and determines the determination result as the control unit 4. Return to

  The voice input means 8 is composed of, for example, an A / D converter, inputs a voice signal generated by the microphone 12 when a person speaks, and converts the voice signal into digital voice data that can be handled by the voice recognition means 3. Convert. The voice data obtained by converting the voice signal by the voice input unit 8 is sent to the voice analysis processing unit 31 of the voice recognition unit 3.

  The screen output means 9 is composed of, for example, a D / A converter, converts the screen output data sent from the output information control means 5 into an analog video signal, and sends it to the monitor 13. As a result, an image composed of characters and pictures corresponding to the screen output data is displayed on the monitor 13. The monitor 13 can be composed of, for example, a liquid crystal display device or a CRT device.

  The audio output means 10 is composed of, for example, a D / A converter, converts the audio output data sent from the output information control means 5 into an analog audio signal, and sends it to the speaker 14. As a result, sound corresponding to the sound output data is output from the speaker 14.

  Next, details of the control means 4 will be described. The control means 4 includes a switching number count processing unit 41, a switching correspondence table 42, a dictionary switching processing unit 43, a recognition engine control processing unit 44, a screen switching processing unit 45, and a recognition result determination processing unit 46.

  The switching count processing unit 41 receives a key event from the manual input means 6 and sends the key code sent simultaneously with the key event to the key code determination means 7. When the key code discriminating means 7 returns that the key code of the recognition start button 15 is returned, the number of times the recognition start button 15 is pressed, that is, the number of times of switching the recognition dictionary is counted. Further, the switching number counting processing unit 41 determines a recognition dictionary to be recognized by referring to the switching correspondence table 42 according to the counted number of switchings, and sends the dictionary number to the dictionary switching processing unit 43. The screen display information corresponding to the dictionary number is sent to the screen switching processing unit 45.

  As shown in FIG. 4, the switching correspondence table 42 stores “recognition target”, “dictionary number”, and “screen display information” in association with the number of times of switching (the number of times the recognition start button 15 is pressed (n)). ing. “Recognition target” indicates whether the number of times of switching corresponds to an address, a facility name, a telephone number, or a song name. The “dictionary number” indicates which of the recognition dictionaries # 1 to # 4 corresponds to the number of times of switching. “Screen display information” indicates which of the numbers 1 to 4 of the output information shown in FIG. The switching correspondence table 42 is referred to by the switching number count processing unit 41 as described above.

  The dictionary switching processing unit 43 generates a dictionary switching request for switching to the recognition dictionary specified by the dictionary number in accordance with the dictionary number sent from the switching number counting processing unit 41, and the recognition dictionary management means 2 Send to. Further, when a dictionary switching request is generated, the dictionary switching processing unit 43 notifies the recognition engine control processing unit 44 to that effect.

  The recognition engine control processing unit 44 requests the speech analysis processing unit 31 of the speech recognition means 3 to start recognition in response to the notification that the dictionary switching request is sent from the dictionary switching processing unit 43. And send a request to stop recognition. In response to the recognition start request and the recognition stop request, the speech recognition means 3 starts or stops the speech recognition processing.

  The screen switching processing unit 45 generates a screen switching request indicating that the screen should be switched to the screen specified by the screen display information according to the screen display information sent from the switching number counting processing unit 41, and outputs information control means Send to 5. Further, the screen switching processing unit 45 sends the recognition vocabulary sent from the recognition result determination processing unit 46 to the output information control means 5.

  The recognition result determination processing unit 46 determines the recognition vocabulary by making a final determination based on the recognition vocabulary sent from the matching processing unit 32 of the speech recognition means 3 as a recognition result and its score. The recognized vocabulary determined by the determination in the recognition result determination processing unit 46 is sent to the screen switching processing unit 45. The recognition vocabulary determined by the recognition result determination processing unit 46 is used in various applications, although a detailed description is omitted.

  Next, the operation of the speech recognition apparatus according to Embodiment 1 of the present invention configured as described above will be described with reference to the flowchart shown in FIG.

  When the speech recognition apparatus is activated, first, the dictionary number n is initialized to “1” (step ST1). Next, it is checked whether or not there is a key input from the remote controller 11 (step ST2). That is, the control unit 4 checks whether a key event has been sent from the manual input unit 6. Here, if it is determined that there is no key input, step ST2 is repeatedly executed while waiting for a key input to be entered, and a standby state is entered.

  When a key is input in the standby state due to repeated execution of step ST2, it is checked whether or not the recognition start button 15 is pressed (step ST3). That is, the switching number count processing unit 41 of the control unit 4 sends the key code sent together with the key event from the manual input unit 6 to the key code determination unit 7. Then, in response to this, it is checked whether or not it is a key event of the recognition start button 15 based on the discrimination result sent back from the key code discrimination means 7.

  If it is determined in step ST3 that the recognition start button 15 is pressed, the recognition dictionary is switched (step ST4). That is, the switching number count processing unit 41 counts up the switching number n, and refers to the switching correspondence table 42 to determine a recognition dictionary to be recognized. Then, the dictionary number of the determined recognition dictionary is sent to the dictionary switching processing unit 43, and screen display information corresponding to the dictionary number is sent to the screen switching processing unit 45. The dictionary switching processing unit 43 generates a dictionary switching request according to the dictionary number and sends it to the recognition dictionary management means 2. The recognition dictionary management means 2 selects one recognition dictionary to be recognized in response to the dictionary switching request, sets the state of the selected recognition dictionary to “valid”, and sets other recognition dictionaries to “invalid”. Set. Since the number n of switching is initially set to “1”, recognition dictionary # 1, that is, an address dictionary is selected.

  Next, switching between screen display and audio output is performed (step ST5). That is, the screen switching processing unit 45 generates a screen switching request according to the screen display information sent from the switching number counting processing unit 41 and sends it to the output information control means 5. Next, screen display and audio output are performed (step ST6). That is, the screen output generation unit 52 of the output information control unit 5 that has received the screen switching request responds to the screen switching request and displays the screen corresponding to the nth recognition dictionary from the output information table 51 (see FIG. 3). The characters and utterance example characters are read to generate screen output data, which is sent to the screen output means 9. Thereby, the screen display characters and the utterance example characters are displayed on the monitor 13. Since the number of switching is initially set to “1” and the address dictionary is selected, a screen for recognizing an address as shown in FIG. In addition, the voice output generation unit 53 of the output information control unit 5 reads the voice guidance characters corresponding to the nth recognition dictionary from the output information table 51 in response to the screen switching request from the screen switching processing unit 45, and generates a voice. Output data is generated and sent to the audio output means 10. Thereby, voice guidance is output from the speaker 14. At first, since the number of switching is set to “1” and the address dictionary is selected, the voice “Please tell me the address” as shown in FIG. 6A is output.

  Next, voice recognition is started (step ST7). That is, the recognition engine control processing unit 44 of the control unit 4 generates a recognition start command when the dictionary switching processing unit 43 is notified that a dictionary switching request has been generated, and the speech analysis processing of the speech recognition unit 3 Send to part 31. Thereby, the voice recognition process is started.

  Next, it is checked whether or not a key input has been made within a certain time (for example, within 1 second) (step ST8). If it is determined that a key input has been made, the dictionary number n is incremented (+1) (step ST9), and then the sequence returns to step ST3. Thereafter, the above-described processing is repeated. Therefore, each time the recognition start button 15 is pressed, the recognition target dictionary is recognized as dictionary # 1 (address dictionary) → recognition dictionary # 2 (facility name dictionary) → recognition dictionary # 3 (phone number dictionary) → recognition dictionary # 4. (Song name dictionary) → Recognition dictionary # 1 (address dictionary)... Further, the screen display on the monitor 13 and the sound output from the speaker 14 are as shown in FIG. 6 (a) → the state shown in FIG. 6 (b) → the state shown in FIG. 6 (c) → FIG. 6 (d). The state shown in FIG. 6A changes cyclically to the state shown in FIG. 6A shows that when the recognition start button 15 is pressed 1, 5, 9,... Times, FIG. 6B shows that the recognition start button 15 is pressed 2, 6, 10,. 6 (c) shows that when the recognition start button 15 is pressed 3, 7, 11,... Times, FIG. 6 (d) shows that the recognition start button 15 is 4, 8, 12,.・ ・ Shows the display screen and audio output status when pressed once.

  If it is determined in step ST8 that no key input has been made within a predetermined time, a speech recognition process is executed (step ST10). That is, the voice input from the microphone 12 is converted into voice data by the voice input unit 8 and sent to the voice analysis processing unit 31. The voice analysis processing unit 31 analyzes the voice data sent from the voice input unit 8 and sends the analysis result to the matching processing unit 32. The matching processing unit 32 executes a matching process for comparing the analysis result sent from the speech analysis processing unit 31 with the vocabulary of the recognition dictionary selected at that time, and recognizes the obtained recognition vocabulary and its score. As a result, the result is sent to the recognition result determination processing unit 46 of the control means 4. The recognition result determination processing unit 46 determines the recognition vocabulary based on the recognition vocabulary and the score sent as the recognition result from the matching processing unit 32 of the speech recognition unit 3 and sends the recognition vocabulary to the screen switching processing unit 45. The screen switching processing unit 45 sends the recognition vocabulary sent from the recognition result determination processing unit 46 to the output information control means 5.

  Next, a recognition result is presented (step ST11). That is, the screen output generation unit 52 of the output information control unit 5 that has received the recognized vocabulary from the screen switching processing unit 45 generates screen output data based on the recognized vocabulary and sends it to the screen output unit 9. Thereby, the character of the recognition result is displayed on the monitor 13. Further, the audio output generation unit 53 of the output information control unit 5 generates audio output data based on the recognized vocabulary from the screen switching processing unit 45 and sends it to the audio output unit 10. Thereby, the recognition result is output from the speaker 14 by voice.

  If it is determined in step ST3 that the recognition start button 15 has not been pressed, the voice recognition process is terminated if the voice recognition process has already been started (step ST13). Processing is executed (step ST14).

  In the speech recognition apparatus according to the first embodiment configured as described above, when the user wants to perform an address search, the user presses the recognition start button 15 once, speaks to the microphone 12, and sets the song title. When searching, the user presses the recognition start button 15 four times and speaks to the microphone 12. Thereby, the target vocabulary can be recognized with a high recognition rate by pressing the recognition start button 15 and once speaking.

  As described above, according to the speech recognition apparatus according to the first embodiment of the present invention, since the dictionary used for speech recognition is selected by pressing the recognition start button 15, the recognition process is performed. The vocabulary can be narrowed down by selecting a certain recognition dictionary, and the target vocabulary can be recognized with a high recognition rate. Simultaneously with the switching of the recognition dictionaries, the switched recognition dictionaries are displayed on the monitor 13 and output by voice, so that a recognizable vocabulary can be presented to the user. In addition, since the recognition dictionary used for speech recognition is configured to share the recognition start button 15 for starting speech recognition and is selected according to the number of times the recognition start button 15 is pressed, the speech recognition apparatus can be simplified. Moreover, it can be configured at low cost. Furthermore, the target vocabulary can be recognized at a high recognition rate by pressing the recognition start button 15 and uttering a continuous word once even for continuous words that are not one sound.

Embodiment 2. FIG.
In the speech recognition apparatus according to Embodiment 2 of the present invention, the speech recognition dictionary 1 is constituted by a plurality of recognition dictionaries classified by the first phonetic character in the vocabulary.

  The configuration of the speech recognition apparatus according to the second embodiment is the same as that of the speech recognition apparatus according to the first embodiment shown in FIG. 1 except for the configuration of the speech recognition dictionary 1. The speech recognition dictionary 1 is made up of ten recognition dictionaries # 1 to # 10 classified according to the 50-sound table rows. Recognition dictionary # 1 is an “a-line dictionary” that includes vocabularies (eg, Enoshima Aquarium) that begin with a-line (a, i, u, e, o), and recognition dictionary # 2 is k-line (ka, ki, k, ke , This) includes a vocabulary beginning with (for example, Kasai Rinkai Park), similarly, recognition dictionary # 3 is "sa line dictionary", recognition dictionary # 4 is "ta line dictionary", recognition dictionary # 5 Is a “na line dictionary”, a recognition dictionary # 6 is a “ha line dictionary”, a recognition dictionary # 7 is a “ma line dictionary”, a recognition dictionary # 8 is a “ya line dictionary”, and a recognition dictionary # 9 is a “la line dictionary”. The recognition dictionary # 10 can be composed of a “wa line dictionary”.

  The operation of the speech recognition apparatus according to the second embodiment configured as described above is that the ten recognition dictionaries # 1 to # 10 constituting the speech recognition dictionary 1 are sequentially switched when the recognition start button 15 is pressed. Except for this, the operation is the same as that of the speech recognition apparatus according to Embodiment 1 described above.

  As described above, according to the speech recognition apparatus according to the second embodiment of the present invention, the recognition dictionary is sequentially switched every time the recognition start button 15 is pressed, and at the same time, the screen of the monitor 13 and the speaker 14 are switched. Is output in accordance with the selected recognition dictionary (for example, “can speak a vocabulary beginning with Aiueo”). Therefore, the user speaks after switching to the recognition dictionary including the vocabulary he / she wants to utter. For example, when the user wants to utter “Kasai Rinkai Park”, the utterance is made after pressing the recognition start button 15 twice because the dictionary is a Ka-line. Thereby, the target vocabulary can be recognized with a high recognition rate by pressing the recognition start button 15 and once speaking.

  In the second embodiment, the speech recognition dictionary 1 is composed of ten recognition dictionaries # 1 to # 10 classified by 50-note table rows. Etc.), 5 recognition dictionaries # 1 to # 5 classified according to the above, 50 recognition dictionaries # 1 to # 50 classified according to each phonetic character, or a plurality of recognitions obtained by grouping arbitrary phonetic characters It can also consist of a dictionary.

Embodiment 3 FIG.
The speech recognition apparatus according to Embodiment 3 of the present invention is such that a specific recognition dictionary can always be set as a recognition target dictionary. As the specific recognition dictionary, for example, a recognition dictionary including a high-priority vocabulary can be assigned.

  The configuration of the speech recognition apparatus according to the third embodiment is the same as that of the speech recognition apparatus according to the first embodiment shown in FIG. 1 except for the stored contents of the management table 21 held in the recognition dictionary management means 2. Is the same.

  FIG. 7 is a diagram showing the configuration of the management table 21 held in the recognition dictionary management means 2. The management table 21 stores a dictionary number (# 1 to # 4), a dictionary name (address, facility name, telephone number, song name) and state (invalid, valid or always valid). Then, similarly to the management table 21 according to the first embodiment, one recognition dictionary to be recognized is selected in response to the dictionary switching request sent from the control unit 4, and the state of the selected recognition dictionary Is set to “valid”, and other recognition dictionaries are set to “invalid”. However, the recognition dictionary set to “always valid” is not changed by a dictionary switching request generated by the normal depression of the recognition start button 15.

  When the desired recognition dictionary is set to “always enabled”, the recognition start button 15 is pressed for a long time while a screen corresponding to the recognition dictionary is displayed. For example, in order to make the song name dictionary always valid, the recognition start button 15 is pressed for a long time while the screen shown in FIG. As a result, the song name dictionary is set to “always valid” and is always set as the recognition target dictionary. Therefore, since the recognition dictionary can be switched while always using the song name dictionary as the recognition target dictionary, a plurality of recognition dictionaries such as a song name dictionary and a facility name dictionary can be simultaneously set as the recognition target dictionary. The recognition dictionary management means 2 sends the contents of the recognition dictionary set to valid and always valid to the voice recognition means 3 when voice recognition is performed.

  Next, the operation of the speech recognition apparatus according to Embodiment 3 of the present invention will be described with reference to the flowchart shown in FIG. The flowchart shown in FIG. 8 is configured by adding steps ST21 and ST22 to the flowchart of the first embodiment shown in FIG. Hereinafter, only the parts different from the flowchart shown in FIG. 5 will be described.

  If it is determined in step ST3 that it is a key event for the recognition start button 15, it is then checked whether or not the recognition start button 15 has been pressed for a long time (step ST21). Specifically, the switching count processing unit 41 measures the time from when a key depression key event is received from the manual input means 6 until the key release key event is received, and when this time is equal to or greater than a predetermined value. It is determined that the button has been pressed for a long time. If it is determined in step ST21 that the recognition start button 15 is pressed for a long time, a process of setting the dictionary to “always valid” is performed (step ST22). In other words, the switching count processing unit 41 sends the dictionary number of the recognition dictionary corresponding to the screen selected at that time to the dictionary switching processing unit 43. The dictionary switching processing unit 43 generates a constantly valid setting request according to this dictionary number and sends it to the recognition dictionary management means 2. The recognition dictionary management unit 2 sets the state of the corresponding recognition dictionary to “always valid” in response to the always valid setting request. If it is determined in step ST21 that the recognition start button 15 is not pressed for a long time, the sequence proceeds to step ST4, and the processing already described with reference to the flowchart shown in FIG. 5 is executed.

  As described above, according to the speech recognition apparatus according to the third embodiment of the present invention, for example, a recognition dictionary including a high-priority vocabulary is set to “always valid” by pressing the recognition start button 15 for a long time. Since other recognition dictionaries can be switched by pressing the recognition start button 15, the range of vocabulary to be recognized can be expanded. Further, since the recognition start button 15 for starting the voice recognition is shared and the selection is made by pressing the recognition start button 15 for a long time, the voice recognition apparatus can be configured easily and inexpensively.

  The recognition dictionary set to be always valid is returned to the invalid state by, for example, pressing the recognition start button 15 again for a long time while the screen corresponding to the recognition dictionary is being output to the monitor 13. Can be configured.

Embodiment 4 FIG.
A speech recognition apparatus according to Embodiment 4 of the present invention is a speech recognition apparatus according to Embodiment 2, in which a vocabulary with a modifier is uttered and speech recognition is performed, and the vocabulary is narrowed down based on the recognition result. It is.

  As described above, the names of a plurality of recognition dictionaries stored in the management table 21 held in the recognition dictionary management unit 2 are displayed on the monitor 13. Therefore, since the user can grasp the classification of the recognition dictionary held by the speech recognition apparatus, the user speaks the recognized vocabulary with the dictionary name added as a modifier, that is, utters the dictionary name and the recognized vocabulary. For example, the utterance vocabulary “Kasai Rinkai Park” is used.

  As shown in FIG. 9, the configuration of the speech recognition apparatus according to the fourth embodiment is that a dictionary selection command is sent from the recognition result determination processing unit 46 of the control means 4 to the recognition dictionary management means 2. This is different from the configuration of the speech recognition apparatus according to the above. The functions of the recognition dictionary management unit 2 and the recognition result determination processing unit 46 are different from those of the speech recognition apparatus according to the second embodiment. Hereinafter, only differences from the speech recognition apparatus according to Embodiment 2 will be described.

  The recognition result determination processing unit 46 and the recognition dictionary management unit 2 of the speech recognition apparatus according to the fourth embodiment are configured by adding the following functions to those of the speech recognition apparatus according to the second embodiment. . That is, when the recognition result sent from the matching processing unit 32 of the speech recognition means 3 is the vocabulary “A line”, “K line”,. Is notified to the recognition dictionary management means 2 to select a recognition dictionary corresponding to. The recognition dictionary management means 2 selects one recognition dictionary to be recognized in accordance with the dictionary selection command sent from the recognition result determination processing unit 46, sets the selected recognition dictionary to be valid, and sets the other Disable the recognition dictionary.

  Next, the operation of the speech recognition apparatus according to Embodiment 4 of the present invention will be described. Assuming that the speech “Ka-Kasai Rinkai Park” is spoken, the speech recognition means 3 first recognizes the vocabulary “Ka-Line” and notifies the recognition result determination processing unit 46 of that. When receiving the notification that the vocabulary “ka line” has been recognized, the recognition result determination processing unit 46 sends a dictionary selection command to the effect that the “ka line” dictionary should be selected to the recognition dictionary management means 2. In response to this dictionary selection command, the recognition dictionary management means 2 selects a recognition dictionary for “K”. Thereby, in the next speech recognition, the recognition dictionary of “K” is used.

  Next, the voice recognition unit 3 recognizes the vocabulary “Kasai Rinkai Park” and notifies the recognition result determination processing unit 46 of the recognized vocabulary and the score. The recognition result determination processing unit 46 determines the recognition vocabulary based on the recognition vocabulary sent from the voice recognition means 3 as the recognition result and its score, and sends the recognition vocabulary to the screen switching processing unit 45. The screen switching processing unit 45 sends the recognition vocabulary sent from the recognition result determination processing unit 46 to the output information control means 5. As a result, the recognition result “Kasai Rinkai Park” is displayed on the monitor 13, and the recognition result “Kasai Rinkai Park” is output from the speaker 14.

  As described above, according to the speech recognition apparatus according to the fourth embodiment of the present invention, the recognition process is executed by narrowing down the dictionary by one utterance of the user, so that a high recognition result can be obtained.

Embodiment 5 FIG.
The speech recognition device according to Embodiment 5 of the present invention is a speech recognition device according to Embodiment 2, in which the speech recognition means 3 narrows the vocabulary to the user when the speech recognition means 3 cannot uniquely identify the recognition result. Is intended to encourage utterances.

  The configuration of the speech recognition apparatus according to the fifth embodiment is the same as the configuration of the speech recognition apparatus according to the fourth embodiment except for the function of the recognition result determination processing unit 46. Hereinafter, only differences from the speech recognition apparatus according to Embodiment 4 will be described.

  That is, the recognition result determination processing unit 46, for example, if the recognition vocabulary sent from the matching processing unit 32 of the speech recognition means 3 and the score thereof do not satisfy a predetermined condition, the score is smaller than a predetermined value. When a plurality of recognized vocabularies are obtained, a message for prompting re-speech is generated and sent to the screen switching processing unit 45. The screen switching processing unit 45 sends the message sent from the recognition result determination processing unit 46 to the output information control means 5. As a result, the character prompting the re-speech is displayed on the monitor 13, and the sound promoting the re-speech is output from the speaker 14. For example, when the user speaks “Akasaka” and the speech recognition means 3 obtains two recognition results “Akasaka” and “Takasaka”, “Ai no Akasaka? Is displayed on the monitor 13 and is output from the speaker by voice.

  In response to this, when the user re-speaks, for example, “Aiue no Akasaka”, the speech recognition means 3 first recognizes the vocabulary “Aiue no Akasaka” and recognizes that fact as a recognition result determination processing unit. 46 is notified. When the recognition result determination processing unit 46 receives a notification that the vocabulary “Aiue no Akasaka” has been recognized, it sends a dictionary selection command to the recognition dictionary management means 2 to select the “A line” dictionary. In response to this dictionary selection command, the recognition dictionary management means 2 selects a recognition dictionary for “K”. As a result, the recognition dictionary “A” is set to be used. Next, the voice recognition means 3 executes the recognition process again using the recognition dictionary of “A”. Thereby, the vocabulary “Akasaka” is obtained as a final recognition result.

  As described above, according to the speech recognition apparatus according to the fifth embodiment of the present invention, because the recognition result using the recognition dictionary selected in response to the message prompting the recurrent speech is acquired, The recognition result can be narrowed down, and the recognition rate can be improved.

It is a block diagram which shows the structure of the speech recognition apparatus which concerns on Embodiment 1 of this invention. It is a figure which shows the memory content of the management table used with the speech recognition apparatus which concerns on Embodiment 1 of this invention. It is a figure which shows the memory content of the output information table used with the speech recognition apparatus which concerns on Embodiment 1 of this invention. It is a figure which shows the memory content of the switching corresponding table used with the speech recognition apparatus which concerns on Embodiment 1 of this invention. It is a flowchart for demonstrating operation | movement of the speech recognition apparatus which concerns on Embodiment 1 of this invention. It is a figure for demonstrating the example of the screen display output by the speech recognition apparatus which concerns on Embodiment 1 of this invention, and an audio | voice output. It is a figure which shows the memory content of the management table used with the speech recognition apparatus which concerns on Embodiment 3 of this invention. It is a flowchart for demonstrating operation | movement of the speech recognition apparatus which concerns on Embodiment 3 of this invention. It is a block diagram which shows the structure of the speech recognition apparatus which concerns on Embodiment 4 of this invention.

Explanation of symbols

  DESCRIPTION OF SYMBOLS 1 Voice recognition dictionary, 2 Recognition dictionary management means, 3 Voice recognition means, 4 Control means, 5 Output information control means, 6 Manual input means, 7 Key code discrimination means, 8 Voice input means, 9 Screen output means (output means) DESCRIPTION OF SYMBOLS 10 Voice output means (output means), 11 Remote control, 12 Microphone, 13 Monitor (output means), 14 Speaker (output means), 15 Recognition start button, 21 Management table, 31 Voice analysis process part, 32 Matching process part, 41 switching frequency count processing unit, 42 switching correspondence table, 43 dictionary switching processing unit, 44 recognition engine control processing unit, 45 screen switching processing unit, 46 recognition result determination processing unit, 51 output information table, 52 screen output generation unit, 53 Audio output generation means.

Claims (6)

Multiple recognition dictionaries for recognizing speech,
A recognition start button for starting speech recognition;
Control means for effectively setting one of the plurality of recognition dictionaries according to the number of times the recognition start button is operated;
Voice recognition means for performing voice recognition using a recognition dictionary effectively set by the control means ,
The control means sets the selected recognition dictionary to be always valid when the recognition start button is continuously pressed for a predetermined time or more with one recognition dictionary selected.
The speech recognition apparatus according to claim 1, wherein the speech recognition means performs speech recognition using a recognition dictionary set to be valid and always valid .   Multiple recognition dictionaries classified by vocabulary type, address dictionary including vocabulary used for address, facility name dictionary including vocabulary used for facility name, phone number dictionary including vocabulary used for phone number The speech recognition apparatus according to claim 1, further comprising a song name dictionary including vocabulary used for music names.   The speech recognition apparatus according to claim 1, wherein the plurality of recognition dictionaries are classified according to the first phonetic character of the vocabulary.   A plurality of recognition dictionaries are classified according to the rows of the 50 syllabary, “A line dictionary”, “K line dictionary”, “Sa line dictionary”, “Ta line dictionary”, “N line dictionary”, “Ha line dictionary”. 4. A speech recognition apparatus according to claim 3, comprising: a “ma line dictionary”, a “ya line dictionary”, a “la line dictionary”, and a “wa line dictionary”. Output means for outputting information representing the classification of a plurality of recognition dictionaries,
The control means effectively sets one of a plurality of recognition dictionaries according to the utterance of the information representing the classification when the utterance of the information representing the classification output to the output means is recognized by the speech recognition means. The speech recognition apparatus according to claim 1. An output means for outputting information representing a message;
When a plurality of recognized vocabulary words are obtained by voice recognition by the voice recognition means, the control means outputs a message prompting the utterance to narrow down the vocabulary to the output means, and responds to the utterance made in response to the message To enable one of multiple recognition dictionaries,
2. The speech recognition apparatus according to claim 1, wherein the speech recognition means performs speech recognition again using a recognition dictionary that is effectively set by the control means.

Images