JPS6318400A - Voice recognition/processing controller - Google Patents

Abstract

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

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a voice recognition processing control device that enables multiple control operations based on multiple voice inputs.

[Prior Art] Usually, in a speech recognition device, one speech recognition section is subordinate to one speech input section. That is, the voice information input into the microphone set as the voice input section is supplied to a specified voice recognition section connected to this microphone, and this recognition section analyzes and identifies the voice information input into the microphone. It is designed to perform processing.

Therefore, this one voice recognition device is used exclusively by one person, and even if various tasks are performed using voice recognition, only one person can perform the tasks using voice recognition. It becomes monopolized and becomes an obstacle to improving work efficiency.

In addition, in the case of 1 where only one voice recognition unit is set,
The processing of input speech is entirely entrusted to that one device, and it is predicted that the speech recognition rate will be unstable. Furthermore, if there is background noise such as an impulse input, it may be difficult to classify it as noise, which may cause a decrease in the recognition rate. It may happen.

[Problems to be Solved by the Invention] This invention has been made in view of the above points, and aims to streamline various tasks using voice input by making it possible to perform tasks using multiple voice inputs in parallel. In addition to making it possible to measure voice input, it also prevents the recognition rate from deteriorating due to ambient noise that exists when human voice input is performed, and also enables the execution of voice recognition operations that are optimal for the content of human voice input. It is an object of the present invention to provide a voice recognition processing control device that effectively improves the reliability and efficiency of work.

[Means for Solving the Problems] That is, the voice recognition processing control device according to the present invention includes a voice input unit including a plurality of microphones, etc., and a plurality of voice input units each independently set and controlled by a central processing unit. It is composed of a plurality of voice recognition sections that are centrally managed and controlled, and each of the plurality of voice input sections and the plurality of voice recognition sections are selectively connected to each other by a switch circuit. This not only allows the voice input section and the voice recognition section to correspond on a 1:1 basis, but also allows one voice input section to correspond to a plurality of voice recognition sections.

[Function] In the voice recognition processing control device as described above, a plurality of voice input units are used independently, and work can be performed in parallel by multiple voice inputs, improving work efficiency. In addition, by using part of a plurality of voice inputs for noise detection, it is also possible to perform a process of removing noise components from the voice input for recognition.

Therefore, a great effect is exhibited in stably improving the recognition rate. It is also possible to use multiple speech recognition units with different functions for each recognition target. For example, by separating the functions for alphabetical characters, numbers, etc., it is possible to effectively improve recognition efficiency. can.

[Embodiments of the invention] Hereinafter, one embodiment of the present invention will be described with reference to the drawings.

FIG. 1 shows its basic configuration, and this voice recognition processing control device includes two microphones, a first and a second microphone (M
ICl, MIC2) 111 and 112 are shown.

For these two microphones Ill, 112,
This device includes two sets of first and second voice recognition units (VRU 1, VRU 2) 121 and 122.
22 respectively, and the above-mentioned Seicrophones 111 and 112
and are selectively connected to each other by a selection switch circuit 13.

The two sets of speech recognition units 121 and 122 are both centrally managed by the central processing unit 14, and although details are not shown, the central processing unit 14 has a large capacity. Equipped with an external storage device, various pattern data for speech recognition processing, software data for specifying recognition contents and executing recognition operations, etc. are supplied to the speech recognition sections 121 and 122 from the central processing section 14 as appropriate. It is now possible to do so. In this case, the central processing unit 14 executes time-sharing processing operations as appropriate;
The plurality of speech recognition units 121 and 122 are configured to be able to perform recognition operations in parallel. The central processing unit 14 supplies a switching command to the selection switch circuit 13 in accordance with the recognition command content of the voice recognition units 121 and 122.

The central processing unit 14 is configured to receive control commands specifying work contents etc. from the keyboard 15 as appropriate.

2 to 5 respectively show the selection switch circuit 13.
In the case of FIG. 2, the first and second two microphones 111 and 112 are connected to the first and second voice recognition units 121, respectively.
and 122 are shown. In addition, in the case of FIG. 3, the voice input from the first microphone lit is transmitted to the first and second voice recognition units 121 and 12.
Input from the second microphone 112 can also be supplied to the first and second speech recognition sections 121 and 1221 in parallel. In the case of FIG. 4, the audio input from the first and second microphones 111 and 112 is
In FIG. 5, the voice input from the first microphone 111 is supplied to the second voice recognition section 122.

Here, the voice recognition method in the voice recognition units 121 and 122 may be fixed in hardware depending on the purpose, but generally speaking, if it is flexibly controlled by software, central processing Various methods can be adopted depending on instructions from the unit 14.

In other words, the audio input section (microphones 111 and 112)
The audio input signal from the first
and is sent to either of the second speech recognition units 121 and 122, where a speech recognition operation is executed. In this case, the switch circuit 13 that distributes the input audio signal is switched and controlled by a command from the central processing unit 14, but the control command for the central processing unit 14 that generates this switching command is issued from the audio input unit. It may also be a voice command. Alternatively, it may be based on a command input from the keyboard 15. Various recognition operation modes are set corresponding to the various states of the selection switch circuit 13 shown in FIGS. 2 to 5''.

To explain each work mode, first, when the switch circuit 13 is set as shown in FIG.
Microphone Ill, 112 which is each audio input section
Furthermore, the voice recognition units 121 and 122 are now set independently, and the voice recognition processing from the first and second microphones 111 and 112 is executed in parallel, and the two parallel recognition processes are performed in parallel. It is possible to process multiple tasks at the same time. In this case, time-sharing processing is executed in the central processing unit 14, and a plurality of processing programs are executed.

Specifically, the first microphone 111 and the voice recognition unit 121 execute the recognition process of the input voice, and the second
Based on the input from the microphone 112, the second speech recognition unit 122 can perform operations such as registration of the speaker's speech pattern, learning, and testing. Also, of course, it allows multiple workers to input a large amount of data in parallel, making it easier to speed up the work and improve efficiency. .

Further, it is possible to set the first microphone lit to perform audio input and the second microphone 112 to input ambient noise at the time of this input. In this case, the second speech recognition m 122 supplied with the input information from the second microphone 112 recognizes not only the steady noise in the environment surrounding the speaker (in this surrounding environment) When impulse-like noise occurs, the presence of this noise can be confirmed. Therefore, from the input audio information from the first microphone ill, the noise component input from the second microphone 112 It is possible to perform processing to invalidate the corresponding signal components, which is particularly effective in improving the speech recognition rate in noisy environments.

Next, when the selection switch circuit 13 is connected to the state shown in FIG. 3, one voice input section exclusively uses a plurality of voice recognition sections.

That is, the voice input from the first microphone 111 is simultaneously supplied to the first and second voice recognition sections 121 and 122. In such a setting mode, the first and second voice recognition units 121 and 122
can be made effective by installing programs with different recognition methods, and the same voice input can be performed by the voice recognition units 121 and 122.
Speech recognition processing is performed using different algorithms. Therefore, the recognition probability of the speech recognition field can be sufficiently improved.

In the case of FIG. 4, the combination state of the voice input section and the voice recognition section can be switched, and a state where only one of them is used can also be set as shown in FIG.

In such a usage state, for example, the first voice recognition unit 121 is equipped with a recognition program that executes recognition operations exclusively for numbers, and the second voice recognition unit 122 is equipped with a recognition program that executes recognition operations exclusively for alphabets. Make sure the program is installed.

Therefore, if such a usage pattern is set, by switching and controlling the switch circuit 13, it becomes possible to use the recognition program that is most suitable for the type of input audio, and the recognition process of the input audio can be executed efficiently. In addition, the recognition rate can be easily improved.

The selection of the work mode as described above may be executed by commands from the keyboard 15 as described above, but it is more effective to execute it by voice commands input by voice.

When performing mode selection using such a voice command, for example, the switch circuit 13 is set to the state shown in FIG. The state is set so that the voice recognition unit 121 of the voice recognition unit 121 can recognize the voice.

That is, as shown in FIG. 6, when the power switch of this system is first turned on in step 101, step 1
At 02, the system is started up and controlled. Next, in step 103, the switch circuit 3 is initially set so that the first microphone 111 is connected to the first voice recognition section 121, and in step 104, the first microphone 111 is connected to the first voice recognition section 121. 121 to load the voice command voice pattern. Then, in step 105, the voice command voice is input to the first microphone, phone 111, and this voice command is recognized by the voice recognition unit 121 of the shield 1, and in step 10B, the voice command is recognized by the central processing unit 121.
At step 107, the content of the cammond is determined, and a switching command is given to the selection switch circuit 13 in accordance with the above determination result.

That is, the state of the switch 13 shown in FIGS. 2 to 5 is commanded and set by voice input, and the central processing unit 14 understands the content of the input voice,
In addition to selecting the state of the switch 13 as described above, the recognition programs of the first and second speech recognition sections 121 and 122 are switched and set. In this case, the voice commands mentioned above can be inputted not only when the system starts up, but also when it is necessary to switch the work mode. The state of the switch circuit 13 as shown in FIG. 2 is set by voice input;
1, a voice pattern for voice command recognition is set.

In the above embodiment, the voice input section and the voice recognition section are shown in two sets each, but it goes without saying that they may be set in a larger number of sets. In addition, the recognition method in the speech recognition unit may be fixed in hardware if particularly high-speed processing is desired, but if more effective recognition processing is desired, each recognition method can be fixed from an external storage device. It is more effective to load the information into a memory such as a RAM set in the speech recognition unit using software, in order to execute various recognition processing operations.

[Effects of the Invention] As described above, according to the speech recognition processing control device according to the present invention, various tasks using multiple speech inputs can be efficiently executed in parallel, and noise components present at the time of input can be separated. etc. can be effectively performed, the speech recognition probability is sufficiently improved, and speech input work can be performed reliably and with high efficiency, especially in noisy environments.

[Brief explanation of drawings]

FIG. 1 is a block diagram illustrating a speech recognition processing control device according to an embodiment of the present invention, and FIGS. 2 to 5 each show an example of the switching setting state of the selection switch circuit used in the above embodiment. FIG. 6 is a flowchart illustrating the voice command input processing state in the above embodiment. ill, 112...Microphone (audio input section)
, 121.122... Voice recognition unit, 13... Selection switch circuit, 14... Central processing unit. Applicant's agent Patent attorney Takeshi Suzue 3.1j.2 Figure 3 Figure 4

Claims (1)

[Scope of Claims] A plurality of voice input units each configured independently to perform voice input; a plurality of voice recognition units each configured independently and executing a voice identification operation based on input voice information; A central processing unit that centrally processes the voice identification operations executed by each of the plurality of voice recognition units in a time-sharing manner; and a central processing unit that is controlled by the central processing unit and performs mutual processing with the plurality of voice input units and the plurality of voice recognition units. and a switch circuit for selectively connecting the plurality of voice input sections, and each of the plurality of voice input sections is selectively connected to one or more of the plurality of voice recognition sections. A voice recognition processing control device characterized by: