JPS6121527A - Voice recognition input device - Google Patents

Abstract

PURPOSE:To decrease the number of times of rejection by collating a pattern of a voice uttered by a talker and a registered pattern and re-registering the inputted pattern in place of the registered pattern when they are dissident for a prescribed number of times. CONSTITUTION:When an ID code is set from a code input section A2 by a ten- key, a recognition mode is set and the registration pattern with respect to the talker is loaded to a main memory D from a storage area of an external memory F. When the talker utters a word toward a voice input section A1, the voice is subjected to analysis and patterning by a recognition processing section E and collated with the registered pattern. When the input pattern is coincident with the registered pattern, a control section C outputs a code of the registered pattern to a main processor G. When the result of collation is dissident and the number of times of rejection reaches a specified value, the word code is inputted and re-registration is displayed. When the talker inputs the word code and utters a required word, the recognizing processing section E applies analysis and patterning to it, the result is stored in a storage area of the main memory D and the re-registration is finished.

Description

[Detailed Description of the Invention] 3.1 Industrial Application Field This invention stores words uttered by a speaker as a registered pattern in a storage unit, and registers words uttered arbitrarily by the same speaker as an input pattern. This invention relates to a speech recognition input device that compares the words with patterns to determine whether they are the same word or not, and if they are the same, inputs the word to the main processing device.

3.2 Prior Art and Its Disadvantages In conventional devices of this kind, the speaker's registered pattern is fixedly stored in the internal storage device, and at the time of input, it is simply checked whether the input pattern matches the registered pattern or not. Only the following judgments are made, and the results are rejected as erroneous recognition or not applicable.
In the event of a misrecognition, the corrective action would be to simply re-enter the word verbally, or the staff member would use a keystroke to output the same word as the uttered word to the main processing unit. Or, every time an attempt is made to input the word ``1~'' and the AA 3B of ``-'', voice input or key operation is required, making the work complicated.

3.3 Objects of the present invention and means for achieving the same In view of the above-mentioned points, the present invention provides that the input 1.
If iji is selected from the registered words and is in the middle 6, as a temporary measure, the input word can be easily re-registered in place of the registration pattern of the lj word that cannot be selected.
The purpose is to minimize the number of wA recognition or rejections, and to minimize the number of re-entries and, in particular, equipment downtime due to rejections.

In order to achieve the above object, the present invention stores a registered pattern for each speaker in an external memory, and upon input of a speaker's ID code, the registered pattern of the speaker is transferred from the external memory to the main memory of a processing device. In addition to loading
A mode setting section that can be set to either recognition mode or registration mode is provided, and a code input section is provided for specifying the registered pattern storage area of the main memo 1). The input pattern obtained by analyzing and patterning the uttered voice is compared with the registered pattern read from the main memory, and if the result is a match, the code of the registered pattern is output to the main processing unit. , If the result is a mismatch, the code of the registered pattern is output to the main processing unit based on the input from the code input section when the number of times of mismatch reaches a predetermined number, and when the registration mode is set, The input pattern obtained by analyzing and patterning the voice uttered by the speaker after the code input from the code input section is re-registered in the storage area specified by the code input in the main memory.

3.4 Examples of the present invention' Next, embodiments of the invention will be described based on the drawings.

FIG. 1 is a block diagram schematically showing the configuration of the present invention, and the input section A includes an audio input section A+ consisting of a microphone that converts audio into an electrical signal and an amplifier that amplifies the electrical signal, and a cord. Input section A2 and selection input section A
It consists of 3. Code input section A1 is the speaker's ID
It is used to set the code or the registered pattern storage area of the main memory D, for example, numerically, and is made up of a numeric keypad or the like. However, the setting of the ID code is not limited to input using the numeric keypad. For example, the ID code may be input by having a card reader read an ID card owned by each speaker.

The display section B is, for example, a CRT (cathode ray tube) liquid crystal display, a plasma display, 1. , ED Display characters, etc.
It is used to display images, etc., and to request selection of an action, promote operation, prompt for confirmation, or indicate the end of an action.

The selection/input section A3 is a transparent touch switch (pressure-sensitive or temperature-sensitive switch) A that is arranged on the display surface of the display section and provides signals with two contradictory meanings.
31. It is made up of A32.

The main memory D stores a control program for the voice recognition input device and a voice recognition/registration program as fixed data, and also temporarily stores a registration pattern that is loaded by inputting a speaker's ID code.

The recognition processing section E includes an analysis circuit that analyzes the audio signal input from the audio input section A, a patterning circuit that forms an input pattern by patterning its characteristics into a pattern similar to the registered pattern, and a main memory that stores the input pattern. It consists of a matching circuit that compares the registered pattern read from D, and outputs the matching result.

F is an external memory provided according to the present invention, and as illustrated in FIG. 2, it is preferable to use a floppy F1 capable of high-speed reading by random access. As schematically shown in FIG. 3, this external memory F stores predetermined ID codes (1, 2 to K) for each speaker identified as the person in charge of this voice recognition input device, and each episode. Coded registered patterns of words uttered by the person are stored. The registration pattern in the external memory can be stored in an external device, or by storing the word registration program in the main memory of this device and setting the mode setting section, which will be described later, to the registration mode. This can also be done by running the program.

Then, when an ID code is input from the code input section A2, a group of address numbers (01, 02, 02,
. . n) and the encoded registration pattern corresponding to the address number are read out and loaded into the main memory D.

The control unit C controls each unit A by executing the control 10 grams described above.
, B, controls the transmission and reception of signals or data between D and F.

ID code input by the code input unit A2 of the input unit A and selection input at a predetermined step by the selection input unit set the recognition mode or registration mode of the voice recognition input device of the present invention, as will be described later. Configures a mode setting section for

The operation of this voice recognition input device having the above configuration will be explained based on the flowchart of FIG. 4 and FIG. 5.

When the N source is turned on and an ID code is set by operating the numeric keypad from the code input section A2 or by reading the ID card, the recognition mode is set for the control section C, and the control section C inputs the ID code from the external memory F. Load the registered pattern registered for this speaker from the storage area specified by the input ID code into the main memory D (pl
). When the loading is completed, display section B displays that voice input is possible for the speaker (p3). Based on this, when the speaker utters the required word toward the voice input section A1 (lea), the voice is analyzed and patterned by the recognition processing section E, and the input pattern is the main one. It is compared with the registered pattern in memory D (p5)
.

If the input pattern matches the registered pattern, that is, if the input word matches the registered word, the control unit C:
Output the code of the registered pattern to the main processing unit @G (
13s), the input processing of this voice recognition input device is ended.

On the other hand, if the result of the verification is a mismatch, it is determined whether the number of rejections, that is, the number of consecutive times n that was determined as not applicable, has reached a predetermined value, for example, 2 (
+17>, V1=2, the display unit B displays that voice input is possible again and prompts the speaker to speak again (p 3 ).

If the number of rejections reaches n=2, display section B will inform the speaker that the code for the word that was determined to be not applicable should be input from code input section A2, and that it should be re-registered. is displayed (p8).

When the speaker follows this and inputs the code of the word from the 5-word input section, the process jumps to the sixth step (p6), where the code is output to the main processing unit G and the desired word is input. In addition, following the input of the code, the fifth
As exemplified in figure a, characters or figures with the same meaning are displayed (, D, i o ) requesting the user to select whether or not to re-register the word that has just been input into the main memory D.

At this time, characters for specifying whether to select re-registration or not are displayed on the same display surface at positions corresponding to the touch switches A 31 + A 32.

If the speaker operates a switch indicating that re-registration is not to be performed, the recognition/registration program that has been executed thus far is terminated.

On the other hand, if the speaker operates the switch to select re-registration, this will cause the mode setting section to select the registration mode.
The word code is set to "code", and characters indicating that the code of the word to be registered should be input from the code input section Δ2 are displayed on the display section B, as illustrated in FIG. 5b (1112).

Based on this, when the speaker inputs ]-, the code just entered is displayed in the code display area of the same screen, and the code corresponding to the touch switch below is displayed, as shown in FIG. 5C. , display a message requesting confirmation of the code.

If you operate the No touch switch, a message will appear asking you to enter the code again. Further, when the OK touch switch is operated, a message indicating that voice input is possible is displayed on the display unit. When the speaker utters the desired word, once the recognition processing section E has completed the analysis and patterning, the control section C stores the input pattern in the storage area of the main memory D designated by the previous input code. Store. This completes the re-registration of the word.

When the input of the word is completed, as illustrated in FIG. 5d, a message indicating the completion is displayed, and the execution of the recognition/registration program is completed.

Therefore, if this speaker inputs the same word from now on,
Speech recognition is performed based on this re-registered registered pattern. Therefore, for example, even if you are unable to speak normally due to a cold or other poor physical condition, and are unable to input as not applicable at the first input, you will be able to input voice from now on by re-registering once. If the user's physical condition has improved when inputting the ID code at a later date, the input can be performed in the same way as normal using the registered pattern in the external memory that is loaded when setting the ID code.

3.5 Effects of the Present Invention As explained above, according to the present invention, a registered pattern for each speaker is stored in an external memory, and by inputting the ID code of the speaker, the registered pattern of the speaker is A voice recognition input device configured to load data from an external memory into the main memory of a processing device, comprising a mode setting section for setting a recognition mode or a registration mode, and a code input for specifying a registered pattern storage area of the main memory. When the recognition mode is set, the input pattern obtained by analyzing and patterning the voice uttered by the speaker is compared with the registered pattern read from the main memory, and if the result is a match, the input pattern is checked. outputs the code of the registered pattern to the main processing unit, and when the result is a mismatch, the code of the registered pattern is outputted to the main processing unit based on the input from the code input section when the number of times of mismatch reaches a predetermined number. When the registration mode is set, the input pattern obtained by analyzing and patterning the voice uttered by the speaker after the code input from the code input section is specified by the code input in the main memory. If the voice of the speaker changes due to poor physical condition, etc., and the input word is not selected from the registered words, temporary measures can be taken. Instead of the registered pattern of the word that cannot be selected, the input word can be easily re-registered to minimize the number of misrecognitions or rejections, and to minimize the number of re-inputs and, in particular, equipment downtime due to rejection. can be kept to a limit.

[Brief explanation of the drawing]

The drawings show one embodiment of the present invention; FIG. 1 is a block diagram showing the basic configuration, FIG. 2 is a front view showing a specific example of the input section and display section, and FIG. 3 is an external view. Fig. 4 is a flowchart showing the flow of operations; Fig. 5 is a schematic diagram showing the relationship between loading of memory and main memory; Fig. 5 is a flowchart showing the flow of operations;
An example of display contents at each step is shown below. Figure 4 Figure 1

Claims (1)

[Claims] Registered patterns for each speaker are stored in an external storage device (hereinafter referred to as external memory), and the speaker's personal identification code (hereinafter referred to as ID code) is input. A voice recognition input device configured to load registered patterns from the external memory to a main memory (hereinafter referred to as main memory) of a processing device, comprising: (a) a mode setting unit for setting a recognition mode or a registration mode; , a code input section for specifying a registered pattern storage area of the main memory, and (b) when setting the recognition mode, the input pattern obtained by analyzing and patterning the voice uttered by the speaker and the main memory are provided. The registered pattern is compared with the registered pattern read from the memory, and if the result is a match, the code of the registered pattern is output to the main processing unit, and if the result is not a match, the number of times the pattern has been repeated reaches a predetermined number. (c) When the registration mode is set, a voice uttered by the speaker after the code is input from the code input section; A voice recognition input device characterized in that an input pattern obtained by analyzing and forming a pattern is registered in a storage area of the main memory designated by the code input.