JPS59226940A - Electronic device containing speech recognizing function - Google Patents

Abstract

PURPOSE:To attain the quick response and to increase the resistance to noise with improvement of reliability for an electronic device containing a speech recognizing function, by recognizing first only a calling word and also obtaining an instruction queuing mode to limit the instruction enable time. CONSTITUTION:The sounds supplied from a terminal 14 are fed to an amplifier 16 via a filter 15 which transmits the sound bands and converted into a digital form by an A/D converter 17 to undergo the data processing 18. When the processing 18 is over, the digital sounds are transferred to a register 19 by a CPU21 and compared with a standard instruction memory 20 by a comparator 22. The result of this comparison is discriminated by the CPU21, and a recognition output 27 is delivered if the degree of coincidence is higher than a certain reference level. A timer 25 which sets the instruction queuing time displays the sound and the light with the output 26 after a calling word is recognized. Thus the instruction queuing state is informed. The operation is invalidated if no instruction is delivered during the display of the sound and light. In addition, a sound eliminating circuit 28 which inverts the phase and eliminates the amplifier output for the sounds of an electronic device itself.

Description

[Detailed Description of the Invention] The present invention relates to an electronic device with a voice recognition response function.
Regarding its composition.

An object of the present invention is to provide an easy-to-use voice recognition response method in the field of relatively small consumer electronic appliances that do not require a high recognition rate.

Traditionally, electronic devices with voice recognition functions were
1 computers, and there are no examples of them being used for television or radio. This is partly because consumer products are expensive, but also because audio equipment is always the source of noise, making it difficult to recognize it.

The present invention proposes a method that can be incorporated into such consumer products.

The present invention will be explained below based on the drawings.

FIG. 1 shows a processing routine for voice recognition responses according to the present invention.

First, input l or co; the first call word is input.

Specifically, it is input as an electrical signal through a microphone. This first address word, for example "television", is compared with reference word 2 in comparator 3. The comparison method is performed by using a well-known speech recognition technique after digitizing an analog speech signal. After comparison, a response is made when it is determined that the word matches the reference word. This involves synthesizing speech from the first reply synthesis memory 5 and performing reply processing 4. For example, the user may respond with "Hi." At the same time, it enters the command standby mode 6, generates a standby display signal 7 indicating that it is waiting for the next command, and opens the gate 8 for receiving the next command to wait.

During this time, you can lower your voice slightly on TV or the like to make it easier to read the commands. To do this, the audio circuit is controlled using a standby display signal. The command standby mode is preliminarily limited to a certain period of time. For example, it is about 15 seconds. The command entered during this period is compared with a specific command word memory 10 by a comparator 9, and when it is recognized, it is synthesized into speech from the second reply synthesis memory 12, and the reply processing 11 is performed again, as well as identification for command execution 13. Generate a signal. For example, if the command is "3 channels", a signal for setting the channel to 3 is generated. In the example of a television, the command words are limited to things like changing channels, turning on/off switches, and raising and lowering the volume, so the number of commands is about 15 at most, and even if it is output as a binary code, 4 bits is sufficient.

A configuration in which the television is operated based on the identification signal is possible using well-known techniques. A feature of the present invention is that when waiting for a command, a display to that effect is made by sound, light, or display, and if the command is not issued while the display is displayed, the command is invalidated. By doing this, you can always concentrate on recognizing only the call word at first, and you can increase the recognition rate.

It is also possible to use a fixed method to recognize predetermined address words and command words.

It would be preferable to limit the word length and type and allow the user to set them.

FIG. 2 is a block diagram of the speech recognition section.

The audio input from 14 is input to the amplifier 16 through a filter 15 that passes the audio band, and then AD converted at 17 and then output to the 1
8, the data is processed. After data processing s ”pu
21 and is transferred to the register 19, and is then transferred to the standard instruction memory 20 and the comparison circuit 22.
It is compared in. The results are determined by cps, and if the degree of matching is above a certain level, a recognition output 27 is output. 23
17. is a response circuit, 24 is a speaker, and 25 is a timer that sets the command waiting time after the address word is recognized.
26U is its output. The timer output can be used to indicate the command standby state by displaying a sound or light. 28
is a built-in sound canceling circuit that inverts the phase of the device's own sound and cancels the amplifier output so that the device does not enter a command standby state due to the sound produced by the device itself.

The electronic device with voice recognition function of the present invention can respond quickly because it first recognizes only the call word9.
It can be said that it is resistant to noise. In addition, by creating an instruction standby mode and restricting the instruction available time, the circuit configuration becomes easier and noise countermeasures improve reliability. Furthermore, by allowing users to register greeting words in their confessions, they can be used in a fun way, and there are fewer opportunities to use shouting words carelessly other than intentional calls, which reduces recognition.

As described above, the electronic device with voice recognition function of the present invention is highly effective as a low-cost and easy-to-use product with voice response.

[Brief explanation of the drawing]

FIG. 1 is a processing routine for a voice recognition response based on the present invention. 1. Input 2. Standard word 3. Comparison 4. Reply processing 5. First reply synthesis memory 6. Waiting for command 7
・Standby display signal 8 inside ・Gate 9 ・Comparator lO−
φ Command word memory 1111・Reply processing 12・Second reply synthesis memory 13・Fist command execution FIG. 2 is a block diagram of the voice recognition part. 14@・Voice human power 15・・Filter 16・・Amplifier 17・・AD conversion circuit 18・・Data processing circuit 19
・-Register 20+10 standard instruction memory 21...
CPU/Fist comparison circuit O/E response circuit...Speaker 5...Timer 26...Timer output N...Recognition output side...Built-in front cancellation circuit and above Applicant Suwa Seikosha Co., Ltd. Agent Patent Attorney Sai Superior Affairs Figure 1

Claims (1)

[Claims] 1. A first recognition response function that recognizes and responds only to a specific call word; a second recognition response function that recognizes and responds to a specific command within a predetermined time after the first recognition response; 1. An electronic device with a voice recognition function, characterized in that it has a recognition response function, and a command standby display means for displaying the time for accepting the specific command with sound, light, a display mark, or the like. 2. Claim 1, characterized in that part or all of the address word and specific command can be initialized by the user.
Electronic equipment with voice recognition function as described in Section 1.