JPH04106596A - Voice recognition and output device - Google Patents

Abstract

PURPOSE:To lighten the voicing load on an operator greatly by inputting only a voice of a key word and then obtaining an output with a large information quantity. CONSTITUTION:A voice recognition part 2 compares a voice which is inputted through a microphone 1 with a standard pattern to recognize the voice. A reader 4 receives the recognition result 3 from the voice recognition part 2 and retrieves a key word stored in a memory 5. The memory 5 is stored with the addresses of voice synthesis patterns stored in a voice synthesis pattern memory 6. Further, a key word A is inputted to the recognition part 2 through the microphone 1. Then the reader 4 retrieves a memory 5 stored with correspondence between key words and the addresses of the respective voice synthesis patterns and reads the address of the voice synthesis pattern corresponding to the key word recognized by the recognition part 2. Then data of the voice synthesis pattern whose address is specified is sent to a voice synthesis part 7 and a synthesized voice 9 of a long sentence is outputted from a speaker.

Description

DETAILED DESCRIPTION OF THE INVENTION (Field of Industrial Application) The present invention relates to a speech recognition output device that has a speech recognition function and outputs information corresponding to the recognition result.

(Prior Art) In the past, speech recognition devices were generally connected to output devices such as word processors and speech synthesis devices, and output in the form of text or speech depending on the content of the recognized speech. .

FIG. 2 is a diagram showing a conventional speech recognition processing flow.

First, when the analog voice signal 2o is input to the voice recognition device 30, it is amplified by the microphone amplifier 21 with an appropriate gain. And after that, analog low pass filter 2
2 performs band limitation to prevent aliasing distortion. The analog audio signal that has undergone the above processing is converted into a digital audio signal by the A/D converter 23. Then, in the next frequency analysis by the bandpass filter 24, it is converted into a spectrum which is the final voice data expression. At this time, the power of the audio signal is also calculated.

As described above, the voice converted into a spectrum is processed by the voice section detector 25 which detects the voice section using power information etc.
The starting and ending points of the true voice section can be found. A matching section 27 matches the spectrum data within the true voice section with a standard pattern 26 prepared in advance. Through this matching, a distance value representing the degree of dissimilarity is calculated for each category. The final recognition result 28 is determined based on the magnitude of the distance value and is output.

Next, the output recognition result is sent to the output device 29 for executing the process intended by the user, and output is performed according to the recognition result. For example, a word processor converts the recognized word into a character string and displays it on the display. Or output the character for each printer. Also,
If it is a voice translator, after translating the recognized word into another language,
Output the machine language using voice translation.

(Problems to be Solved by the Invention) However, the above-described conventional technology has the following problems.

That is, in an output device using voice recognition, the amount of information of input voice and output is equivalent. In other words, in order to obtain an output containing a large amount of information, such as a long sentence or guidance, the amount of vocalization must be increased, which poses a problem in that the burden on the speaker increases.

The present invention has been made with attention to the above points, and in order to eliminate the problem that the input voice and output of an output device using voice recognition have the same amount of information, and the burden of speaking is large, only keywords can be used. It is an object of the present invention to provide a device which is capable of outputting a large amount of information in an associative manner by inputting information, and is easy to operate with a low burden of vocalization.

(Means for Solving the Problems) A speech recognition output device of the present invention includes a speech recognition unit that recognizes speech, and a memory that stores information that associates keywords with output contents, The present invention is characterized in that the output content corresponding to the selected keyword is output from the memory.

(Operation) In the speech recognition output device of the present invention, speech is input using keywords. Therefore, the operator does not need to fully utter the output contents such as sentences and guidance. As a result, the operability of the device can be improved.

(Embodiment) FIG. 1 is a block diagram showing an embodiment of the speech recognition output device of the present invention.

The illustrated device includes a microphone 1, a voice recognition section 2, and a reader 4.
, a memory 5 , a speech synthesis pattern memory 6 , a speech synthesis section 7 , and a speaker 8 .

The microphone 1 is used by the operator to input audio information.

The speech recognition unit 2 recognizes speech by comparing the speech input through the microphone 1 with a standard pattern.

The reader 4 receives the recognition result 3 from the speech recognition unit 2, and
Search for keywords stored in memory 5.

The memory 5 stores speech synthesis pattern addresses corresponding to keywords. This speech synthesis pattern address is the address of the speech synthesis pattern stored in the speech synthesis pattern memory 6.

The speech synthesis pattern memory 6 is a memory that stores speech synthesis patterns that are output contents.

The speech synthesis unit 7 synthesizes speech from the speech pattern.

The speaker 8 outputs the voice synthesized by the voice synthesizer 7.

Next, the operation of the above-described device will be explained.

First, a keyword A is input to the speech recognition section 2 through the microphone 1. Next, the reader 4 searches the memory 5 in which the correspondence between the keyword and the address of each speech synthesis pattern is written, and reads out the address of the speech synthesis pattern corresponding to the keyword recognized by the speech recognition section 2. Then, the data of the addressed speech synthesis pattern is sent to the speech synthesis section, and a long sentence of synthesized speech 9 is output from the speaker.

Next, a second example will be described.

In the first embodiment, only keywords were input, but in the second embodiment, in addition to keywords, situations in which the keywords are used can also be input. This makes it possible to output a single keyword in multiple ways depending on the situation.

FIG. 3 is a block diagram showing a second embodiment of the present invention.

The illustrated device includes a speech recognition section 11 and a knowledge database 1.
5 and a speech synthesis section 17.

The speech recognition unit 11 recognizes the input speech by comparing it with a standard pattern.

The knowledge database 15 consists of a memory that stores speech synthesis patterns corresponding to keywords, and a reader that searches for the keywords stored in the memory.

The speech synthesis unit 17 synthesizes speech from the speech pattern.

Next, the operation of the above-described device will be explained.

First, the voice ``coffee'' is input to the voice recognition unit 11 and recognized. Next, “Shopping” 12, “Restaurant”
13. Select one from the three situations 14: "Break". Note that any method may be used to select this situation. For example, push-button input, voice input, etc. can be considered. Furthermore, the number of times a situation is input is not limited to once, but may be input multiple times in order to deal with more complex situations.
If a keyword is entered after the situation has been entered, the knowledge database will be searched. Through this search, the address of the optimal speech synthesis pattern corresponding to each situation for the input keyword is found. Next, the searched speech synthesis pattern 16 is sent to the speech synthesis section 17 and output from the speaker. For example, when you select ``Shopping,'' the output is ``How much is the coffee?'' When you select ``Restaurant,'' ``Coffee, please.'' When you select ``Break,'' the output is,
``Let's go get coffee.''

As described above, in the second embodiment, it is possible to output one keyword according to the situation.

Further, as an advanced version of the second embodiment, a method of outputting a longer sentence by inputting a plurality of keywords may be considered.

Note that although a speech synthesis translator is used as an example of the output device shown here, various other applications such as a word processor, a display device, etc. can be considered.

(Effects of the Invention) As explained above, according to the speech recognition output device of the present invention, it is possible to obtain an output with a large amount of information just by inputting the voice of a keyword, which greatly reduces the burden on the operator. can be reduced to

Furthermore, by providing a judgment function according to the situation,
Sentences can be estimated and output from the voice of the keyword. As a result, it is possible to expand the scope of use of the device for one keyword.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing a first embodiment of the present invention, FIG. 2 is a block diagram showing the configuration of a conventional speech recognition device, and FIG. 3 is a block diagram showing a second embodiment of the present invention. ■...Microphone, 2...Speech recognition unit, 4...League, 5...Memory, 6...Speech synthesis pattern memory,
7...Speech synthesis section, 8...Speaker.

Claims (1)

[Scope of Claims] 1. A voice recognition unit that recognizes voice, and a memory that stores information associating keywords with output content, and output content corresponding to the keywords recognized by the voice recognition unit is stored in the memory. A speech recognition output device characterized by outputting from. 2. The speech recognition output device according to claim 1, wherein the semantic information of the keyword is searched from a knowledge database, and the output content is estimated from the keyword according to the specification of the situation.