JPS61151706A - Voice recognition control system - Google Patents

Abstract

PURPOSE:To respond to an automatic device according to a conservation voice by applying control corresponding to a key word and a retrigger in response to recognition when the retrigger word is recognized after a pre-registered key word is recognized. CONSTITUTION:A portable mobile station 1 and a robot 10 are communicated by ratio waves with frequencies f1, f2. A voice recognizer 12 of the robot 10 recognizes the received ratio wave, a conversation control section 13 controls a voice synthesizer 14 in response to the recognized output and the response voice is outputted. Further, when various commands are given from a keyboard 24, a main control section 21 and the section 13 respond thereto and a desired control condition is set. When the keyboard 24 commands the preparation of registration of the key word and the trigger word, the device 13 responds thereto and controls the device 14 to output a synthesized voice representing that the registration preparation is set, controls the device 12 to set registration preparation. Then each word is transmitted from a microphone 3a to registrate it and each word of the same uttering personnel is recognized individually afterward.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a method for recognizing speech and performing predetermined control.

[Conventional technology]

Currently, various speech recognition methods have been developed and are gradually becoming more general-purpose, but all of them only recognize words that are made up of a relatively small number of sounds, and only recognize a series of conversational sounds. Recognizing this and performing control accordingly is limited by the configuration of the device, making it impossible to put it to practical use.

[Problem that the invention seeks to solve]

This means that it is impossible to make automatic devices such as robots perform conversational voice recognition and take actions accordingly, and it is impossible for automatic devices to respond according to conversational voice commands. This has resulted in a problem in which the automatic device's response situation is unnatural.

[Means for solving problems]

In order to solve the above-mentioned problem, the present invention is constructed by the following means.

That is, a keyword and a trigger word that are registered in advance for speech recognition are determined, and when the keyword is recognized and a trigger word is subsequently recognized, the control corresponding to the keyword and trigger word is performed in accordance with the recognition of the trigger word. It is said that

[For production]

Therefore, if a keyword is first inserted into a series of conversational commands, and then a trigger word is inserted at the final point in time, the control corresponding to the keyword and trigger word will be performed as needed. , a situation arises in which the automatic device continuously recognizes the conversation and executes the commands according to its termination.

[Implementation row]

Hereinafter, the present invention will be explained in detail with reference to figures showing implementation sequences.

FIG. 2 is a block diagram showing the configuration. A mobile station 1 such as a portable radio is configured with a transmitter/receiver (hereinafter referred to as TRx) 2 and a headset 3, and is picked up by a microphone 3a of the headset 3. Audio is TRX
2 as a radio wave with a frequency fl, and is received by the TRX 11 on the robot 10 side, and the reception output of this is a voice recognition device (hereinafter referred to as V(J) 1).
2, speech recognition is performed here, and this recognition output is given to the dialogue control unit (hereinafter referred to as 0CT) 13, and in accordance with this, the CCT 13 controls the speech synthesis device (hereinafter referred to as VSZ) 141. and TR the future response voice.
X11 and amplifier (hereinafter referred to as PA) 15-\ given, TR
Frequency f! via X11 While transmitting by radio waves,
The speaker 16 is driven through the PA 15.

Therefore, when a predetermined sound is given from the microphone 3a, a corresponding synthesized sound is emitted from the speaker 16, received by the TRX2, and can be heard from the receiver 3bK of the headset 3.

In addition, from the CCT 13, a signal corresponding to the voice recognition is sent via the bus to the main control unit (hereinafter referred to as MCT) 21, where judgments are made to control each unit, and the display unit (hereinafter referred to as DP) 22 and a drive circuit (hereinafter referred to as DRY) 23 for each motor that drives each part, control signals are sent from a keyboard (hereinafter referred to as KB) 24 using a numeric keypad. When various direct instructions are given, MCT21, CCT
13 responds and the desired control conditions are set.

Therefore, if you instruct the preparation of keyword and trigger word registration by operating KB24, this will allow you to
The T13 responds and controls the VSZ14 to send out a synthesized voice indicating that the registration preparation state for each word has been set, and also controls the VCE12 to individually record each word from the microphone 3a in order to set these registration preparations. By transmitting the words to V(J12), each word is registered, and from this point on, each word of the same speaker is individually recognized.

FIG. 3 is a diagram showing the recognition status of a keyword (hereinafter referred to as KW) 31 and a trigger word (hereinafter referred to as TV) 32.
Following the previous statement 33, 1 For example, 0.3 to 1.0 s e c
After setting the space time ts, KW31 is given,
After this, the intermediate word 34 is given through the space time ts, and after the space time ts, the TV 32
is finally given, the preceding word 33 and the intermediate word 34 are ignored, first, KW31 is recognized, then TV32 is recognized, and according to the logical product of both recognition results, it is determined to correspond to KW31 and TV32. Control is started based on the time of recognition by the TV 32.

Therefore, if you keep it, you can register it for voice recognition as "KV/31.1 Give me" as TV32.
When you send the message "Robot-kun, please introduce yourselves to everyone and give it to me" from the microphone 3a, it will recognize "everyone" and "taidai" in sequence, and in response to "taidai", for example, it will lower its head before running. The robot rotates once, and the speaker 16 makes a sound such as "I'm a robot, ○○ Kuntesu," and controls for greeting and self-introduction are executed.

The CCT 13 and MCT 21 each include a processor such as a microprocessor, a memory, etc., and the processor executes instructions in the memory and performs the above-mentioned controls while accessing predetermined data to the memory.

FIG. 1 is a 70-chart of the control status by the processor in MCT21, and the signal indicating recognition of KW31 is C
When given from CT13, this routine returns '5TAR
T' L, 1 standby indicator light/blinking I in DP 22 indicating waiting for the next voice, and
1 program number display to be executed' 101 iDP22
8 of the given voice.
The result is 1 tomare? '102 N (No) and 'TV? “If it is N of 103, then the timer configured in the processor will elapse the time %T of, say, 58 ec?
"While 104 is N, repeat steps 101 and after, and as step 104 becomes Y (YES), 'EX'
Return to a separate main routine via IT".

In contrast to the above, if step 103 is Y, 1 standby indicator light is turned off.
Execute the program corresponding to 32 and 1. Move to 112,'
The end of the program? #113 N and the recognition result of the given voice is 1 tomare? Step 112 and subsequent steps are repeated on the assumption that "114 is ON," and if step 113 or 114 is Y, the process returns to the main routine via vAEXIT.

Note that the space time tS in FIG. 3 is, for example, CCT1
3.

Therefore, if TW32 is recognized after being recognized by Kw311, step 112 is immediately executed, and 1 Tomare is recognized after recognizing KW31.
The routine shown in FIG. 1 is interrupted, and the same happens after the recognition of TW32, and while it is possible to arbitrarily prevent the occurrence of unexpected conditions, the routine of FIG.
The same thing applies if it cannot be recognized.

However, in Figure 1, unnecessary steps may be omitted depending on the conditions, and the configuration in Figure 2 can be arbitrarily selected according to the situation, and in Figure 3, the same can be done even if there are multiple KW31s. Various modifications are possible, such as.

〔Effect of the invention〕

As is clear from the above description, according to the present invention, control can be performed in response to conversational commands without a particularly complicated configuration, and remarkable effects can be obtained in humanoid-like automatic devices such as exhibition robots.

[Brief explanation of drawings]

The figures show an embodiment of the present invention; FIG. 1 is a flowchart of the control situation, FIG. 2 is a block diagram, and FIG. 3 is a diagram showing the speech recognition situation. 2.11----- TRX (transmitting/receiving), 3a. Fist microphone, 12 fist...VCE ('t? u recognition device), 13... CCT (dialogue control unit), 14...
Φ...VSZ (speech synthesizer), 16...speaker, 21...MCT (main control unit), 22...D
P (display section), 23... DRV (drive circuit), 24
・Fist...KB (keyboard). Patent applicant: Yamatake Honeywell Co., Ltd. Agent: Masaki Yamakawa (2 people) Figure 1

Claims (1)

[Claims] A keyword and a trigger word registered in advance for voice recognition are determined, and when the keyword is recognized and the trigger word is recognized after this, the control corresponding to the keyword and the trigger word is performed according to the recognition of the trigger word. A voice recognition control method characterized by: