JPS63304297A - Voice recognition equipment - Google Patents

Abstract

(57) [Summary] This bulletin contains application data before electronic filing, so abstract data is not recorded.

Description

DETAILED DESCRIPTION OF THE INVENTION TECHNICAL FIELD The present invention relates to a speech recognition device for telephones, which allows telephone calls to be made automatically based on the results of speech recognition.

Moxibustion unremoved In recent years, research on speech recognition has been active and attracting attention.

At present, word speech recognition devices have already been put into practical use.

Speech recognition devices include speaker-independent recognition methods that can recognize anyone's voice, and specific-speaker methods that register the user's voice in advance.Specific-speaker methods are advantageous in terms of recognition rate and number of words that can be recognized. .

The use of such voice recognition devices in telephones has already been considered (hereinafter referred to as voice telephones).

For example, it is also shown in Japanese Utility Model Application No. 61-104500. When used for telephone calls, it is common to use a voice recognition device for a specific speaker, since the number and party to be registered differ depending on the user.The above example of use has a voice recognition device inside, rather than a general telephone.

It recognizes the voice spoken through the mouthpiece and automatically oscillates the other party's dial according to the result. This device is believed to be useful for people with physical disabilities, especially those who are visually impaired. However, for these people, it is not easy to enter a telephone number by dialing along with voice when registering, and conventional devices are inconvenient.

In addition, this method requires a dedicated voice phone machine to automatically oscillate the dial of the other party based on the recognition result1, making it impossible to carry around.

The present invention was made in view of the above five drawbacks,
You can register by just saying the number without having to turn or press the phone dial.

The purpose was to provide a device that could be used as a voice telephone simply by attaching it to a regular telephone.

1--In order to achieve the above-mentioned object, the present invention includes a part for inputting voice, an analysis part for analyzing data collected by the input part, and holding data analyzed by the analysis part. A voice recognition device is equipped with a numeric voice pattern storage section in which numeric voice patterns are registered in advance, and a numeric voice pattern storage section is used to register voice patterns of digits in advance. and numeric sounds are inputted from the audio input section, converted into a characteristic pattern by the analysis section, and then compared with the numeric audio pattern stored in the storage section to find a code or symbol string corresponding to the pattern that obtained the maximum similarity. The pattern corresponding to the pattern to be registered is stored in memory, and when an unknown input voice is input, the voice is converted into a characteristic pattern by the analysis unit, and then compared with the registered pattern to obtain the maximum similarity. Transferring a code or symbol string corresponding to the pattern to the telephone, or having a numeric voice pattern storage section in which numeric voice patterns are registered in advance, and inputting the voice to be registered and the numeric voice from the voice input section. 1. After being converted into a characteristic pattern by the analysis unit, the frequency sound corresponding to the pattern that obtained the maximum similarity by comparing with the numerical sound pattern stored in the storage unit is stored in correspondence with the pattern to be registered. Then, when an unknown input voice is input, the voice is converted into a characteristic pattern by the analysis unit, and then the frequency sound corresponding to the pattern that has obtained the maximum similarity by comparing it with the registered pattern is transferred to the telephone set. It is characterized by this. The following will explain one embodiment of the present invention.

FIG. 1 is a block diagram for explaining one embodiment of the present invention, in which 1 is a voice input section, 2 is a voice detector, 3 is a switch, 4 is a voice standard pattern, and 5 is a numeric standard pattern. ,6
, 7 is a matching unit, 8 and 9 are maximum similarity calculation units, 10 is a numerical code, 11 is a code selection unit, and 12 is a transfer unit. A voice recognition device is equipped with an analysis section that analyzes the data analyzed by the analysis section, a section that holds the data analyzed by the analysis section, a collation section that matches the data, and a section that transfers the signal to the telephone. It has a numeric voice pattern storage section in which patterns are registered in advance, and the voice to be registered and the numeric voice are inputted from the voice input section, and after being converted into a characteristic pattern by the analysis section, the numeric voice is stored in the storage section. A code or a symbol string corresponding to the pattern that obtained the maximum similarity by matching the pattern is stored in correspondence with the pattern to be registered, and when an unknown input voice is input, the voice is transmitted to the analysis section. After converting the pattern into a characteristic pattern, the code or symbol string corresponding to the pattern that obtained the maximum similarity by comparing it with the registered pattern is transferred to the telephone set.

In Fig. 1, 1 is a voice detector as a voice input section, which is a sound 11/electrical converter such as a pickup or a microphone, and 2 is an analysis section that converts the voice detected by the voice detector into feature quantities. The type of quantity is not particularly limited; for example, if a spectrum is used, it can be realized by a group of bandpass filters. Further, although not shown in the figure, it is effective to provide a voice section detection section and extract only necessary signals. Reference numeral 3 denotes a switch, which allows switching between voice registration mode and recognition mode. Reference numerals 4 and 5 are storage parts that hold registered standard patterns of speech, and in particular, 5 is a part that stores standard patterns of numbers, but this need not be provided separately from part 4 that stores standard patterns of speech. 6 and 7 are the parts that compare and match the standard voice pattern and the input voice pattern. There are several known methods for matching, and the ones described here are does not depend on the matching method, so any means may be used.This is also functionally divided into two separate parts and displayed, but either one may have both functions.

Reference numeral 8.9 is a part that selects the voice pattern that has the maximum degree of similarity based on the results of the matching performed by the matching units 6 and 7. This also allows the matching section to perform both functions. Depending on the method of matching, the minimum distance may be selected instead of the maximum similarity. 10 is a part for associating the numeric code corresponding to the recognition result determined by the maximum similarity calculation unit 8 with the voice recognition pattern, and 11 is a dial code corresponding to the recognition result determined by the maximum similarity calculation unit 9. 12 is a part that transfers the dial code to the telephone set.

Next, how to use it will be explained. The user first flips the switch to the number voice registration side and registers the number voice, but this is done by saying the numbers from 0 to 9. At this time, guidance is provided using synthesized voices etc. to determine the voice to be uttered. It would be better if you let the speaker know.

When the number registration is complete, turn switch 3 to the name registration side, and say the name of the person you want to call followed by the phone number of the person you want to call into the mouthpiece. The voice is analyzed, the person's name is stored in the voice pattern storage section, and the number is compared with a standard voice number pattern in the matching section, and the one with the highest similarity is selected as the recognition result. A numerical code corresponding to the recognition result is selected and registered in a standard pattern in correspondence with the person's name.

When all registration is complete, turn switch 3 to the recognition side,
Speak the name of the person you want to call into the mouthpiece. The voice is analyzed and compared with the standard pattern of one person's name, and the one with the highest similarity is selected as the recognition result. A numeric code corresponding to the recognition result is selected, converted into a dial code, and transferred to the F4 handset.

When using voice recognition, the recognition rate is not high enough, so it is desirable to always show the recognition results to the user and operate the system only when the results are recognized as correct.

Further, considering the device to be portable, the following configuration can be adopted. This is because button-type telephones have a combination of specific frequency sounds! This is done with attention to the fact that it represents the episode number.

FIG. 2 is a block diagram for explaining another embodiment of the present invention. In FIG. 1, the same parts as in the embodiment shown in FIG. Reference numbers are provided.
1 is a voice input section, 2 is an analysis section, 3 is a mode changeover switch, 4 is a storage section for standard patterns, 5 is a section for storing standard patterns of numbers, 6 and 7 are for storing standard voice patterns and input voice patterns. The parts for comparison and matching, 8 and 9, are the parts that select the voice pattern that has the maximum similarity based on the results of the matching in the matching section, and 13 is determined by the maximum similarity calculating section 8! ! ! The part 11 converts the numbers of the recognition results into frequency sounds related to the recognized relationship and matches them with the standard speech pattern. Also, 11 selects the frequency sounds corresponding to the recognition results determined by the maximum similarity calculation unit 9. Some parts may have oscillators, or only certain frequency sounds may be recorded.

12 is the part that transfers the dial code to the 'W1\8 machine. Therefore, this embodiment has a part for inputting audio,
an analysis section that analyzes the data collected by the input section; a section that holds the data analyzed by the analysis section;
A voice recognition device that is equipped with a collation unit that collates data and a unit that transfers signals to a telephone set has a numeric voice pattern storage unit that stores numeric voice patterns in advance, and allows the voice to be registered and the numeric voice to be registered. After inputting from the voice input unit and converting it into a characteristic pattern by the analysis unit, the frequency sound corresponding to the pattern that obtained the maximum similarity by comparing it with the numerical voice pattern stored in the storage unit is added to the pattern to be registered. When an unknown input voice is input, the voice is converted into a characteristic pattern by the analysis unit, and then matched with the registered pattern to generate a frequency sound corresponding to the pattern that obtained the maximum similarity. Fl! The information is transferred to the handset.

Next, how to use the embodiment shown in FIG. 2 will be described. The user first performs numeric and voice registration in the same manner as in the previous embodiment.To register the name of the person he or she wants to call, flip the switch 3 to the name registration side and enter the name and then the phone number of the other person into the mouthpiece. emanate towards. The speech is analyzed, the person's name is stored in the speech pattern storage section, and the number is compared with a standard speech number pattern in the matching section, and the one with the highest similarity is selected as the recognition result. Based on the recognition results, a corresponding frequency sound is selected and registered in a standard pattern in correspondence with a person's name. When all registration is complete, turn switch 3 to the recognition side and say the name of the person you want to call into the mouthpiece. The voice is analyzed and compared with the standard human name pattern, and the frequency sound corresponding to the recognition result with the highest similarity is selected and transmitted to the telephone.

Note that although the above explanation did not discuss the connection of the acoustic transducer in detail, this is not something that is specifically specified.
- For example, you can use something like an acoustic coupler, with the audio input section 1 set at the sending port and the transfer section 12 set at the receiving port. This is because you need a way to be able to have a conversation while connected to the acoustic coupler as soon as you take off the telephone receiver.

It is desirable to be able to use a regular 11-phone device by simply attaching a pickup to the 11-phone device.

As is clear from the above explanation, according to the present invention, registration can be made by simply saying the number without turning or pressing the telephone dial, and moreover, voice calls can be made by simply attaching it to a regular telephone. It can be done.

[Brief explanation of the drawing]

FIG. 1 and FIG. 2 are configuration diagrams for explaining the present invention in detail, respectively. DESCRIPTION OF SYMBOLS 1... Voice input part, 2... Voice detector, 3... Switch, 4... Voice standard pattern, 5... Number standard pattern, 6.7... Verification part, 8, 9 ...Maximum similarity calculation unit. 10...Numeric code, 11...Code selection section, 12
... Transfer section, 13... Frequency selection section.

Claims (2)

[Claims] (1) A part for inputting voice, an analysis part for analyzing the data collected by the input part, a part for holding the data analyzed by the analysis part, a collation part for collating the data, and a telephone. A voice recognition device equipped with a signal transfer section includes a numeric voice pattern storage section in which numeric voice patterns are registered in advance, the voice to be registered and the numeric voice are inputted from the voice input section, and the numeric voice pattern is inputted to the analysis section. After converting it into a characteristic pattern by , when an unknown input voice is input, the voice is converted into a characteristic pattern by the analysis unit, and then the code or symbol string corresponding to the pattern that has obtained the maximum similarity by comparing it with the registered pattern is transferred to the telephone set. A speech recognition device characterized by: (2) A part for inputting voice, an analysis part for analyzing the data collected by the input part, a part for holding the data analyzed by the analysis part, a collation part for collating the data, A voice recognition device equipped with a signal transfer section includes a numeric voice pattern storage section in which numeric voice patterns are registered in advance, the voice to be registered and the numeric voice are inputted from the voice input section, and the numeric voice pattern is inputted to the analysis section. After converting it into a characteristic pattern by When an input voice is input, the voice is converted into a characteristic pattern by the analysis unit, and then compared with the registered pattern, and a frequency sound corresponding to the pattern that has obtained the maximum similarity is transferred to the telephone set. voice recognition device.