JPS6041863A - Pushphone voice input device - Google Patents

Abstract

PURPOSE:To simplify the sequence by providing a code section to both voice recognition section and DTMF ( dual tone multi-frequency) so as to eliminate the inquiry of succession/stop and its response. CONSTITUTION:When a voice recognition section 5 outputs any of 0-11 as the result of recognition and the DTMF decode section 5 outputs the presence of an input signal, a microprocessor 11 selects an output of a decode section 6. When the recognition section 5 outputs any of 12-99 and the decode section 6 outputs the presence of input signal, the processor 11 urges re-input but when the decode section 6 outputs the absence of input signal, the processor 11 selects an output of the recognition section 5. When the recognition section 6 outputs any of 0-11 and the decode section 6 outputs the absence of input signal, the processor 11 gives a command to a voice output section 7 that a voice message urging the re-input is to be transmitted.

Description

DETAILED DESCRIPTION OF THE INVENTION [Technical Field of the Invention] The present invention relates to a butshuphone voice input device that allows specific information to be input into a system by voice from a butshuphone telephone.

[Conventional technology and problems]

FIG. 1 is a diagram showing an overview of a conventional button-phone voice input device, and FIG. 2 is a diagram illustrating a sequence during voice input. In Fig. 1, -1 is a butshuhon telephone, 2 is a voice recognition unit, and 3 is a DTMF (Dual Tone Mu
ltjfrequency) code section, 4 indicates a changeover switch, respectively. If the subscriber wants to perform voice recognition, the subscriber calls the voice recognition system. When a response was sent from the voice recognition device, Mr. Toguchi pressed a button on his phone and entered a number, for example, #nnn. #RunR indicates the start of the voice input request. Voice recognition unit f) f (For li, selector switch 4YD
Fold it to the TMF decoding section 3 side.

Waiting for DTMF signal to be sent, signal #1
When ``■'' is sent, the vDTMF decoding unit 3 analyzes it, learns that voice input is requested, and sends a message from the voice output unit (not shown) to subscriber C ("Please speak"). A voice message is sent, and then the changeover switch 4 is turned to the side of the voice recognition section 2.When the person making the million-mouth purchase utters "desk," this voice signal is input to the voice recognition section 2vc, and the recognition result is processed through a process not shown. After reading the voice recognition dictionary, the voice recognition device sends a voice message to the subscriber saying ``I want to continue, but do you want to stop?'' To continue voice input, the 7JO user presses the button on the handset phone - for example, inputs #m.The voice recognition device receives this signal #m. The DTMF decoder 3 analyzes it, and, knowing that the subscriber is requested to continue voice input, sends a voice meso message asking the subscriber to "please speak" and turns the changeover switch 4 to the side of the voice recognition unit 2. .When the subscriber receives a voice message asking for ``please call,'' he/she utters, for example, ``window.'' This voice signal is input to the voice recognition section 2, and the recognition result is input to a processing device (not shown). .

The conventional voice input sequence is performed as described above, but when using voice input, if you can indicate the beginning and end of the voice input, it is considered to be a light minute, and the voice recognition device says ``Do you want to continue?''"Do you want to stop?" and the response to this are considered unnecessary. If continuation/correction pauses and responses to these are present in the voice input sequence, the voice input sequence becomes unnecessarily complicated.

[Purpose of the invention]

This is what I learned from the above discussion.

11 kat voice recognition IJ' through the discovery bond
It is an object of the present invention to provide a voice input device for a phone that can be used to input voice signals such as 71 to the J sound, and to perform voice input in the case of Z line 5 and the blank ';+: fiil in the case of -.

[Structure of the invention]

[Thus, the invented pushphone 71 voice input device is connected to the pushphone telephone and the above-mentioned pushphone telephone via the telephone line 7, and is connected to the above-mentioned pushphone 1.
(a) a two-frequency combined signal decoder for analyzing the two-frequency combined signal generated when the pushbutton of the handset is pressed; Speech recognition section in which standard patterns are stored; 'F Y R speech recognition section; analysis results of the above two-frequency combined signal decoding section; and recognition results of the above speech recognition section are provided. That is.

[Embodiments of the Invention] Hereinafter, explanation will be given with reference to the following.

Figure 3 is a block diagram of the 19th embodiment of the present invention;
Figure 2j' + 51st (according to Chigi's IC Voice Input Sequence 1 Description: This is a no-ro diagram.

In Fig. 3, 6 is a voice recognition unit, 6 is a DTM,
F'tecode part-7 is voice recognition part-8 and 9 are registers,
]-0 indicates a selector, and 11 indicates a micro processor.

The speech recognition unit 5 has a speech recognition dictionary (1/2) in which support patterns for each category of speech are stored. , the pushbutton signal that is generated when the pushbutton of a pushbutton telephone is pressed λt is also a voice mark pattern in the case ``Kosa21.'' For example, for the standard audio pattern U of the push button signal, the 0th and 11th numbers are divided by y1! :) Jl, So2
tzo'fLO~9, -#:, *corresponds to iC. i
7C/', - in's voice standard pattern C・τ ka [6X 12) If the 990 number is 1τl it:
Fp 0 voice identification capital 5 ba, , each is expressed by a separate connection number, identification result number stone 1/jis 19 envy set 1-ru. D
T 34 FTE code tl+ 6 analyzes ~ push button signal 7, and the analysis result rP: registers 8 (tr-se,). If there is data in the second star 8 or 9, there will be an interrupt for the first micropro Nijinoza 11; I''L
stone. Interruption) Shisura Kurotoichi Microglossena 1
1. Read the contents of registers 8 and 9 through the selector 10. It is determined whether the voice signal uttered by the subscriber is sent from the selector 10 or the push button signal is sent. do.

The feature of the present invention is that the voice recognition section and the DTMF decoding section are arranged side by side, and the voice dictionary of the voice recognition section has push button signals 0 to 9.
．． :#:', *There are 12 standard horns corresponding to *.

This enables highly reliable reception. That is, the 2 mm component constituting the push button signal is within the voice band 1c, and there is a small probability that the DTMF decoding section is operated and output by the voice. f The voice recognition unit operates and outputs an output by the push button signal with a very high probability.

Therefore, there is a fairly high probability that output from both the speech recognition section and the DTMF code section will occur, and the DTM
Since the operation of the F decoding section is also taken into consideration, it is not possible to uniformly receive the push button signal, and re-input is required at a high frequency of ρ.

In the present invention, the IC is configured so that the operation of the voice/recognition section by the push button signal also operates normally. In that case, the recognition number corresponding to the pushbutton signal is output. However, there may be erroneous recognition due to voice, or there may be erroneous recognition between the pushbutton signal and the DTMF. A decoding section is also required.

If received according to the selection criteria shown in Figure 4, the probability that a push button signal will be erroneously recognized as a recognition number 12 to 99 is the probability that a voice recognition unit that does not have a standard pattern corresponding to the push button signal will operate on the push button signal. Since the size is much smaller, highly reliable reception becomes possible.

As shown in FIG. 4, the voice recognition unit 5 outputs any one of numbers 0 to 11 as a recognition result.

D'l'MF decoding unit 5 outputs input signal present L r
, -, the microprocessor 11 selects the output of the DTMF decoder 6. The speech recognition unit 5 outputs any one of the 12th to 99th numbers as a recognition result -DT
When the MF code section 6 outputs an input signal,
The microprocessor 11 prompts for re-input 2. Voice recognition section 5
outputs any one of the 12th to 99th fragrances as a recognition result, and when the DTMF decoder 6 outputs no input signal, the bar microphone o processor 11 selects the output of the voice recognition unit 5. When the voice recognition section 5 outputs a number of numbers 0 to 11 as recognition results and the DTMF decoding section 6 outputs no input signal, the microprocessor 11 outputs a number from number 0 to number 11 as the recognition result. Instruct them to send a voice message urging them to re-enter □. Upon receiving this command, the voice output section 7 outputs the voice mesomesage that is to be re-inputted.

FIG. 5 is a diagram illustrating a sequence Z of voice input according to the present invention. If the subscriber wants to perform a voice call, the subscriber makes a call to the voice recognition device. When a response is sent from the voice recognition device, the 210 people press the button on their handset phone + enter #nnnnyx. When the voice recognition device side 1 recognizes the sound phone signals #y+ and rL, -
A voice message is sent to the subscriber requesting, ``Please speak.'' Upon receiving this voice message, the subscriber will
Say it out loud. This voice signal is sent to the voice recognition device,
Voice input is performed. When the voice input process is completed, the voice recognition @ sender sends a message to the voice message subscriber side saying, "Please say 7." When the subscriber receives this message, he or she utters, for example, "window." This voice signal is sent to the voice recognition device. If it is desired to change such a sequence, it is possible to send a predetermined pushphone signal #mmmq in response to "Please speak". The same process is repeated thereafter.

〔Effect of the invention〕

As is clear from the above description, the present invention allows voice input to be performed without unnecessary effort and time.

[Brief explanation of the drawing]

Fig. 1 is a diagram showing the outline 1 of the conventional voice input system, Fig. 2 is a diagram showing the sequence of voice input, Fig. 3 is a block diagram of one embodiment of the present invention, and Fig. 4 is a diagram showing the voice input system. FIG. 5 is a diagram illustrating an algorithm for selecting either the output of the recognition section or the output of the DTMF code section. 5... Voice recognition unit, 6... DTM, F decoding unit, 7... Audio output unit, 8... Register 1o... Selector, 11... Microprocessor. Patent Applicant: Fujitsu Co., Ltd. Agent Patent Attorney Kyotani Four Parts Total 1 (2) -f Z Kasukone 3 Diagram■ 1'11'41 Kyouo S Diagram

Claims (1)

[Claims] a two-frequency combined signal decoder connected to the push-button telephone via a telephone line and configured to analyze a two-frequency combined signal generated when a pushbutton of the push-button telephone is pressed; - A voice recognition section connected to the phone telephone via the telephone line and having a voice recognition dictionary storing the standard pattern of the two-frequency combination signal, and an analysis result of the two-frequency combination signal decoding section and the voice recognition What is claimed is: 1. A button voice input device characterized by comprising a processor and a processor for reading the recognition results of the parts.