JPH0332810B2 - - Google Patents

Description

[Detailed description of the invention]

[Industrial Application Field] The present invention relates to an interface circuit equipped with a voice recognition device that recognizes voice and outputs an electrical signal. In particular, the present invention is utilized as an interface between a computer device and a terminal. [Prior Art] In general, a speech recognition device extracts sections and characteristic patterns of input speech by analyzing the input speech by frequency band using a bandpass filter (BPF) or by cepstrum analysis. It calculates the similarity between the feature patterns of a finite number of words to be recognized that are stored in advance in the memory, and outputs the word with the feature pattern that is most similar to the input speech to an external host computer as a recognition result. . The speech recognition device also receives a recognition operation control command from an external computer in order to start the recognition operation. This is because the host side can give instructions midway in cases such as when the recognition target word changes or when the recognition method is different between discrete recognition and continuous recognition. As a way to connect such a speech recognition device and a host computer
Cable connections using RS-232C interfaces are often used. Usually, as a host computer, at least one PC or minicomputer is used.
Equipped with an RS-232C interface,
If the voice recognition device is also equipped with an RS-232C interface, it is possible to connect most personal computers and minicomputers to a host computer, at least physically. Conventional speech recognition equipment uses a host computer to
After converting the serial data sent through the RS-232 interface into parallel data, the commands were decoded and various recognition operations were performed based on the results. Additionally, the recognition results were output to the host computer using the same RS-232C interface as above. [Problems to be solved by the invention] In the conventional method as described above, when creating a business application program on a host computer, the port number for connecting the voice recognition device, which is a data input means, and the recognition You must understand everything from the program sequence after receiving the results, from recognition control commands to the speech recognition device being used to the output results, such as processing when the recognition result is "religious" (that is, it does not resemble any word). . This has resulted in the drawback that it takes a lot of time to create a program. In particular, even if the same speech recognition device is used, if the application program has a different data input processing method, the entire program must be rewritten, which increases program creation time and cost. This was also a factor that hindered the spread of voice recognition devices. Furthermore, even if a word is difficult to recognize with the speech recognition device, the application program has no choice but to retry the speech recognition device several times and perform error processing if the word is not correctly recognized. The present invention solves the above-mentioned conventional problems, and provides a speech recognition control interface circuit which has the function of a speech recognition section in a conventional speech recognition device and which places almost no burden on the application program of a host computer. The purpose is to provide. [Means for Solving the Problems] The present invention has a function of accepting recognition operation control command data from the outside, recognizes input speech as a word, and generates predetermined words corresponding to the recognized word as a recognition result. It has a voice recognition unit that outputs data, a storage unit that stores one or more types of recognition operation control command data, and two pairs of input/output ports that can transmit and receive character data. a first switch that outputs input data from one of the output ports and output data of the voice recognition section to the other of the input/output ports; When the comparison result between the comparison section that performs the comparison with the data and this comparison section do not match, output the output data of one of the comparison sections and the output data of the voice recognition section to one of the input/output ports. The second switching device is a component, and when the comparison result of the comparison section matches, the other output data is output to the voice recognition section. [Function] The signal is sent to the host computer. The output of the host computer is displayed on the CRT display of the operating terminal. However, when a specific predetermined word is sent out from the keyboard, it is recognized by the comparator and sent to the speech recognition section. On the other hand, when a control word is input from the microphone to the voice recognition section, it is converted into a character word and sent to the host computer, and the response from the host computer is displayed on the CRT of the operating terminal. [Embodiment] FIG. 1 is a block diagram of an apparatus according to an embodiment of the present invention. FIG. 2 is a diagram showing the connection relationship between the device of this embodiment, an operating terminal, and a host computer.
That is, the device of the present invention is an interface circuit inserted between a host computer and an operating terminal (CRT and keyboard), which passes signals from the operating terminal to the host computer, and passes signals from the host computer to the operating terminal. . Furthermore, the device of the present invention is equipped with a microphone, and when a control word is input by voice through the microphone, the translated character word is input into the host computer, and the response is displayed on the CRT of the operating terminal. In FIG. 1, the receiving section output of the input/output port pair 1 using the RS-232C interface is connected to the input of the switch 11 of this speech recognition device, and the transmitting section input is as follows.
It is coupled to the output of switch 12. The output of the receiving section of the input/output port pair 2 through the RS-232C interface is input to the switch 12 and the voice recognition section 5 via the comparator 3 coupled to the storage section 4. One output of the speech recognition section 5 is input to the switching devices 11 and 12, respectively. Signal 101,1 of speech recognition unit 5
The 02 and 103 outputs are input to a switch 11, a switch 12, and a comparator 3, respectively. The output of the microphone 6 is input to the speech recognition section 5 via an amplifier 7. For the input/output port pair 1 and 2 to which 8-bit character data is input serially, the data transmission procedure with the outside is set to start-stop synchronization, and the transfer speed is also set to match the outside. In the initial state, the comparator 3 directly transfers the data from the receiving section of the input/output port 2 to the voice recognition section 5.
The distributor 13 is controlled so as to pass the data to. Further, the switchers 11 and 12 output the output data of the speech recognition section 5 in the initial state. Therefore, in the initial state, that is, the operation start signal 101 from the new speech recognition unit 5
Until "ON" is applied to 103, the same operation as a conventional external interface is performed. When a recognition operation command is given to the voice recognition section 5 from the receiving section of the input/output port 2, the voice recognition section 5 turns on the operation start signals 101 to 103 and simultaneously starts the voice recognition operation itself. This recognition operation is a discrete recognition mode given as an initial value, and after outputting the recognition result, the process moves on to the next recognition operation. In the subsequent state, if the data received from the input/output port 2 is other than the data stored in the storage unit 4, the comparator 3 controls the distributor 13 and the data received at the input/output port 2 is switched. The signal is sent to the transmitting section of the input/output port 1 through the transmitter 12. Furthermore, data received from input/output port 1 is sent to the transmitting section of input/output port 2. When output data is output from the voice recognition section 5, the signals 101 and 102 are turned OFF, the data is sent to the transmitting sections of the input/output ports 1 and 2, and when the transmission is finished, the signals 101 and 102 are turned OFF again. Furthermore, when the received data from the input/output port 2 matches any of the data stored in the storage section 4, the comparator 3 passes this input data only to the speech recognition section 5, so that the distributor 13 control. In this embodiment, there are four types of received data as shown in the table. VS is the CTL-S code on the keyboard, VQ is the same CTL-Q code, and VD is the same
The CTL-D code and VC also correspond to the CTL-C code, and the voice recognition unit uses data VS to interrupt processing.
Processing is restarted with data VQ, discrete recognition data is started with data VD, and continuous recognition is started with data VC. Once this data is received for data VD and VC, the same recognition operation is repeated until the next different data VD or VC data is received. Normally, input/output ports 1 and 2 operate as if they are directly connected, and when a recognition result of voice input is obtained, it is output to the two ports. The speech recognition unit 5 includes a pattern storage unit 5a that stores feature patterns of all recognition target words, and a word-corresponding output data storage unit that indicates what to output as recognition result data to the outside for each word. 5
b is built in, and after performing recognition processing on the input speech and determining the corresponding word, it converts it into word-corresponding output data and outputs it. The contents of this word-corresponding output data are made in advance to be the same as the words and character strings for control or data to the host computer 1 using the speech recognition section 5.

〔Effect of the invention〕

The advantages of the present invention are (a) With the keyboard, one key input is one data (usually one character).For example, if the control word to the host computer is 6 characters such as "FORMAT", 6 key inputs are required. However, if you use voice recognition, you can output the "FORMAT" character string as a result with one utterance, so if you need more than a few characters as a control word, use this voice recognition to speed up data input. b) Even if the recognition result is incorrect,
Since the results can be recognized immediately on the CRT, corrections can be made using the keyboard. (c) Even if a voice recognition device is introduced into an existing system for data input, the existing system's software does not require any changes, so the system For example, no software costs are incurred for reconfiguration. As described above, by using the voice recognition control interface circuit according to the present invention, it is possible to easily construct a combined keyboard input and voice input system that has a significantly improved processing speed compared to the conventional data input only from the keyboard KB. Further, the present invention
This has the effect that the contents of the storage section described in the embodiment can be freely expanded.

[Brief explanation of drawings]

FIG. 1 is a block diagram of an apparatus according to an embodiment of the present invention. FIG. 2 is a block diagram of a system using an apparatus according to an embodiment of the present invention. 1...RS-232C input/output port, 2...RS-
232C input/output port, 3...Comparison section, 4...Storage section, 5...Speech recognition section, 6...Microphone, 7...Amplifier, 5a...Pattern storage section, 5b...Word corresponding output data storage section .

Claims (1)

[Scope of Claims] 1. A speech recognition unit 5 that recognizes input speech as words and outputs the recognition results as predetermined data corresponding to words; and a speech recognition unit 5 that is connected to a host computer and is capable of transmitting and receiving character data. In the voice recognition interface circuit, the first input/output port 1 is connected to an operation terminal and is capable of transmitting/receiving character data. a first switching device 11 having a received signal as one input, an output signal of the voice recognition section as the other input, and an output connected to the transmission section of the second input/output port; and an output of the voice recognition section. A second switch 12 whose input is a signal and whose output is connected to the transmitting section of the first input/output port, and whose input is the signal from the receiving section of the second input/output port, a distributor 13 that distributes a signal to the other input of the switch and the voice recognition section; a storage section 4 in which one or more data for control commands of the voice recognition section are stored; and the above-mentioned data stored in this storage section. a comparator 3 for comparing the data for the control command and the data arriving at the receiving section of the second input/output port; , means for controlling the first switch, the second switch, and the distributor, and the distributor controls the voice recognition unit when there is a match between the comparison results of the comparator. Regardless, the input of the distributor is connected to the voice recognition section. A voice recognition control interface circuit characterized by: