JPS60120400A - Voice recognition equipment - Google Patents

Abstract

(57) [Summary] This bulletin contains application data before electronic filing, so abstract data is not recorded.

Description

DETAILED DESCRIPTION OF THE INVENTION [Technical Field of the Invention] The present invention relates to a speech recognition device, and particularly to a speech recognition device that performs recognition after registering a speaker's speech pattern in advance.

[Prior art and problems]

FIG. 1 is a diagram showing an example of the configuration of a voice recognition device, and FIG. 2 is a diagram showing an example of the configuration of a handy panel.

In the figure, 1 is the controller, 2 is the display, and 3
is a keyboard, 4 is a voice recognition device knit, 5 is a microphone, 6
61 is a forward key, 62 is a return key, 63 is a re-registration key, and 64 is a pause key.

As shown in Fig. 1, the conventional speaker-specific speech recognition device is
A voice recognition unit 4 recognizes the voice input from the microphone 5 and notifies the result, a keyboard 3 that is merged with the output, and a handy keyboard 3 for giving voice-specific instructions.
It is composed of a panel 6, a display 2 for interaction with an operator, and a controller 1 for controlling input and output thereof.

As shown in FIG. 2, the handy panel 6 has a forward key 61 and a return key 62 for forwarding and reversing prompts, and for erasing a dictionary created as a target for voice matching based on individual voice features. a re-registration key 63 for re-registering;
It also has a pause key 64 for cutting off voice input and returning it to a state where it can be input again.

Next, an overview of voice registration will be explained. First, when the voice registration process is started, a prompt instructing to speak is displayed on the display 2, and the operator speaks according to the prompt. If registration is successful, a prompt for the next word will be displayed, and the process will continue in the same way. If the registration is not performed normally, the operator presses the re-registration key 63 on the handy panel 6 to erase the created dictionary and re-voice. When a word whose voice has already been registered is re-uttered, a process called a learning process is performed to correct the existing dictionary so that it can respond to variations in pronunciation.

Furthermore, if there are similar words that are close to each other in the dictionaries, a process is also performed to correct the dictionary for the similar words. If the operator needs to interrupt voice input midway through, use the handy panel 6.
Multiple inputs are cut off by pressing the pause key 64. Furthermore, by pressing the pause key 64 again, the state returns to the state where multiple inputs are possible.

The conventional registration method described above requires the preparation of special hardware called a handy panel, and there are no particular restrictions on the use of keys on the handy panel. There is a problem in that the similar word dictionary may be destroyed due to repeated corrections of the similar word dictionary.

[Purpose of the invention]

The present invention is based on the above considerations, and provides a voice recognition device that can perform voice registration processing efficiently and effectively without the need to prepare special hardware such as a handy panel. The purpose is to provide

[Structure of the invention]

To this end, the speech recognition device of the present invention is a speech recognition device equipped with a speech recognition unit that performs speech recognition, a display, a keyboard, and a controller that controls overall input/output. , during voice registration processing to create voice features, the controller uses function keys to send and return prompts for voice registration processing, pause, reregister,
In addition to defining functions such as termination, the initial registration in the voice registration process is given priority over the learning process that corrects/adds registered dictionaries, and the results and status of the learning process and recognition process,
The present invention is characterized in that it is configured to display a command input field on a display.

[Embodiments of the invention]

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

FIG. 3 is a diagram showing the configuration of one embodiment of the present invention, FIG. 4 is a diagram showing a screen image of a display to which the present invention is applied,
FIG. 5 is a diagram showing an example of definition of function keys on a keyboard to which the present invention is applied. In Figure 3,
Items 1 to 5 correspond to those shown in FIG.

As shown in FIG. 3, the present invention eliminates the conventional handy panel and provides the function keys of the keyboard 3 with the function. Since a function key can normally be freely defined by a user, there is no problem even if it is uniquely determined in the registration mode. function·
FIG. 4 shows an example of keno definition. In the example shown in FIG. 4, function key A1 is the pause key;
A3 is defined as a prompt return key, A4 is defined as a prompt send key, B1 is defined as an end key, and B2 is defined as a reregistration key. The end key here is a key that has the function of ending the registration process.

Furthermore, in the present invention, learning processing is enabled only after registration processing has been completed for all words, thereby maintaining uniform dictionary quality. In other words, a high recognition rate can be obtained by focusing on specific words consisting of the following words, but the recognition rate decreases for words that have been reversely modified.This is because the dictionary is not distorted by allowing learning to occur partially when registration is completed. Do it like this.

Table 1 shows state transitions during registration by the specific speaker speech recognition device of the present invention, and Table 2 similarly shows state transitions during learning.

The present invention is realized by restricting key operations depending on the mode, as shown in Table 1 and Table 2 [Table 1: At the time of registration].

Next, a fifth image of the voice registration screen to which the present invention is applied is shown.
This will be explained with reference to the figures. In FIG. 5, the utterance level indicates whether the operator's utterance is too loud or too soft, and the registration status indicates to the operator how much registration has been completed relative to the total number of words. Also, similar utterances are
Inform the operator that there is a word (pronounced similarly) that is close to the registered word, and instruct him or her to change the pronunciation.

The degree of completion indicates the degree of completion of the dictionary, and the candidate is used to display a nearby candidate when there is a nearby candidate during learning. Further, since it is difficult to move to an arbitrary word position by simply sending or returning a prompt, a command is used in a command input field for moving a prompt word.

FIG. 6 is a diagram illustrating the flow of processing by the controller in registration mode. The controller has means for assigning meaning to the function keys. For example, it is a utility function. Specifically, when a function key is pressed, a character string corresponding to the function key is notified to the controller. This controller is
As mentioned above, instead of the conventional handy panel, the function keys on the keyboard have similar functions. The flow of processing in the registration mode will be described below with reference to FIG.

■ Check whether all words have been registered. If 0, the learning process is performed, and if the answer is No, the process (■) is performed.

■ Show the next spoken prompt on the display. Next, perform the process (■).

■ Determine whether a utterance was made or a function key was pressed.

If the utterance has been made, the uttered word is registered and the process returns to step 2, and if the function key has been pressed, the step 2 is performed.

■ Check whether there is a re-registration instruction.

In the case of Yes, the process ``■'' is performed, and in the case of No, the process ``■'' is performed.

■ Clear the contents of registered dictionaries. Next, return to the process of ■.

■ Check whether there is an instruction to send the prompt.

In the case of Yes, the process ``■'' is performed, and in the case of No, the process ``■'' is performed.

■ Check whether the operation is possible in response to the prompt sending instruction.

In the case of Yes, the process ``■'' is performed, and in the case of No, the process returns to the process ``■''.

■ Advance the voice prompt pointer to the next point.

Next, return to the process of ■.

■ Check whether the prompt is returned or not.

In the case of Yes, the process of [phase] is performed, and in the case of No, the process of @ is performed.

[Phase] Check whether the action is possible in response to the return instruction from the prompt.

If Yes, process 0 is performed, and if No, return to process ■.

■ Move the vocal prompt pointer back. Next, return to the process of ■.

@ Check whether it is a termination instruction.

If Yes, process O is performed, and if No, process [phase] is performed.

@Perform Bose processing. Next, return to the process of ■.

■ Perform termination processing.

As mentioned above, a specific function basically consists of
When a key is pressed, the processing functions in the controller include determining whether to enable or disable the function of the function key and limiting the next possible operation.

〔Effect of the invention〕

As is clear from the above description, according to the present invention, in a specific speaker speech recognition device, the speech registration process, which is an important element that influences the recognition rate, is performed using special hardware such as a one-day panel. If you prepare, you will be able to carry out the process efficiently and effectively.

[Brief explanation of drawings]

FIG. 1 is a diagram showing an example of the configuration of a voice recognition device, FIG. 2 is a diagram showing an example of the configuration of a handy panel, FIG. 3 is a diagram showing an example of the configuration of an embodiment of the present invention, and FIG. A diagram showing a screen image of a display to which the present invention is applied; FIG. 5 is a diagram showing an example of definition of function keys on a keyboard to which the present invention is applied; FIG. 6 explains the flow of processing by the controller in registration mode. It is a diagram. 1... Controller, 2... Display, 3...
・Keyboard, 4... Voice recognition unit, 5... Microphone, 6... Handy panel, 61... Forward J) key, 62... Return key, 63... Re-registration key, 64
... Pose Ki Patent Applicant Fujitsu Ltd. Representative Patent Attorney Kyo Tani 4 Section 1 Figure I 2 m)X3ffi -'f 4 Leap)! ■ Entokukai Showa GO-120400 (5) 6 Uedai ■ Yes Kubeichi υ history (sato) ■ Nozomi p prompt display ■ Ax P meter K FK (歭7 false key) ■・□.7 ■ Dictionary Liguria ■ ■ Send 7 Fight Heavy B work possible?

Claims (1)

[Claims] A voice recognition unit that performs voice recognition, a display,
In a speech recognition device equipped with a keyboard and a controller that controls overall input/output, the keyboard is provided with function keys, and the controller registers speech to the function keys in speech registration processing for creating speech features. Send and return prompts for processing,
In addition to defining functions such as pause, re-registration, and termination, the initial registration in the voice registration process is prioritized over the learning process that corrects/adds registered dictionaries, and the results, status, and commands of the learning process and recognition process are defined. A speech recognition device characterized in that it is configured to perform a process of displaying an input field on a display.