JPS60146399A - Voice remote controller - 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] Industrial Field of Application The present invention utilizes telephone lines to control tape recorders for answering machines, home air conditioners, and has recently been used in condominiums and installed in various locations in the home. The present invention relates to a voice remote control device that can be connected to devices that centrally manage crime prevention and disaster prevention using sensors, and can also be used as a monitoring device that monitors the current state of disaster prevention through a telephone at a remote location.

Configuration of conventional example and its problems Conventionally, the above-mentioned remote control device was developed, for example, in
As described in Japanese Patent Laid-open No. 41915 and Japanese Utility Model Application No. 61-101716, there are devices that have a unique oscillator and control the transmission of oscillating sound from the transmitter of a telephone handset. As described in the issue, the main structure was to use a voice recognition device to determine the voice of the owner of an answering machine, etc. based on the voice of a specific speaker, and control the tape recorder of the answering machine, telephone, etc. .

However, in the above conventional configuration, if a specific oscillator is used, this oscillator must always be present, and even if a speech recognition device is used, current speech recognition technology is difficult to register a specific speaker. The voice for registration and the voice for recognition are subject to various frequency characteristics and level fluctuations due to the telephone set and the telephone line, so it has not necessarily been possible to put it into practical use. There wasn't. Also, when recognizing speech, use the word to be recognized.
Despite the fact that the number had to be reduced, and it became impossible to identify the speaker, people other than the owner used it.
Tape recorders and the like did not operate easily, or malfunctions occurred due to crosstalk in the audio recorded on the tape recorder.

The causes of these malfunctions will be explained in detail below with reference to the drawings.

FIG. 1 is a block diagram of a conventional speech recognition device. 1st
In the figure, 1 is an input terminal for inputting a voice signal, 2 is a voice analysis means, 3 is a data normalization means, and 4 is a switching means for switching between a voice registration mode and a voice recognizer mode (in FIG. 6 is a standard audio pattern recording means, 6 is an audio pattern matching means, 7 is a determining means, and 8 is an equipment control signal generating means.

The operation of the conventional device configured as described above will be described below. First, the switching means 4 is set to the voice registration mode as shown in FIG. Compress to a few bits. This process is performed during the audio section, and the time series.

The frequency characteristics (hereinafter referred to as audio pattern) are inputted as a standard audio pattern into the standard audio pattern recording means. After inputting the speech patterns of all words to be speech recognized into the standard speech pattern recording means 5, the mode is switched to speech recognition mode. 7C, After performing the above processing on the unknown input audio,
A speech pattern is extracted and matched with the data in the standard speech pattern recording device 5 by the speech pattern matching means 6, all similar speech patterns are extracted, and the effectiveness is judged by the similarity determining means 7. When the similarity is high, The device control signal generating means 8 is activated.

However, when the above-mentioned conventional voice recognition device is used in a telephone line, the following drawbacks occur.

(Li) In the telephone line, including line loss and exchange loss, level fluctuations of up to 30 dB occur due to line coupling, and the voice analysis means 2 cannot accurately detect voice sections.

(2) Due to path loss, exchange loss, etc., the voice passband changes, the voice pattern generated by the data normalization means 3 is deformed, and the degree of similarity decreases.

(3) Since the telephone line is a public line, anyone who knows the telephone number can control the device over the phone, and for safety and security purposes, a function must be provided to identify the speaker.

Purpose of the Invention The present invention corrects the transmission characteristics of a communication line such as a telephone line using voice characteristics, and performs voice recognition over the telephone line by applying hierarchical structure processing using voice passwords to maintain confidentiality. It is an object of the present invention to provide a voice remote control device that can perform voice remote control of a bicycle regardless of frequency characteristic fluctuations and level fluctuations.

Structure of the Invention In order to achieve the above object, the voice remote control device of the present invention has the following features:
a voice analysis means for analyzing the input voice in-band signal;
A transfer characteristic correction means which inputs the correction amount determined by the transfer characteristic difference extraction means, a data normalization means which normalizes the data determined by the voice analysis means, a standard voice pattern matching means which registers data at the time of registration, and an input at the time of recognition. A speech recognition device comprising: a pattern matching means for matching a speech pattern with a standard speech pattern in the standard speech pattern memory; and a means for determining the degree of similarity between speech inputs at the time of registration and during recognition based on the matching output of the pattern matching means. The device is configured to be controlled by an audio band signal using ff17, thereby making it possible to correct the transfer characteristics of a communication line such as a telephone line.

DESCRIPTION OF EMBODIMENTS FIG. 2 is a block diagram of a voice remote control device according to an embodiment of the present invention.

In FIG. 2, reference numeral 9 denotes a voice analysis means that analyzes the characteristics of the voice using a plurality of band-pass filters and orthogonal transformations such as Ohrsch, Hadamard, and Fourier transforms, using changes in the amplitude of the voice as spectrum information such as frequency. 10 normalizes the time-series data of the analyzed audio signal, suppresses the amplitude information of each spectrum to a small amount of variation, suppresses the variation in the volume of the speaker's sound when speech is generated, and further compresses it to a lower number of bits. data normalization means for dividing speech patterns into symbols, 11
is a standard sound pattern memory in which the password normalized by the data normalization means 10 at the time of registration, and the symbol strings of word sound patterns such as "fast forward" and "rewind" and the symbol strings of short ka-ga sounds are registered. , 12 is a pattern matching means for comparing the input speech pattern at the time of recognition with the standard speech pattern recorded in the standard speech pattern memory 11 at the time of registration. Reference numeral 13 denotes a determining means for determining the degree of similarity based on the result obtained by the pattern matching means 12. The determining means 13 sets a threshold for the degree of similarity, and when the degree of similarity is greater than the threshold (i.e., the standard voice device control signal generating means 1 for generating signals for controlling a tape recorder or a voice synthesis output signal (if the pattern is considered to be the same as the pattern)
4, and if the word is a foreign word, no control signal is output to the device control signal generating means 14 and the device is not controlled. Reference numeral 16 denotes energy extraction means for calculating the maximum or average energy of a signal within the audio band such as the first password at the time of voice registration, and reference numeral 16 averages the first password in the time axis direction for each of the plurality of analysis channels determined by the voice analysis means 9. 17 is an energy extraction means 16, a recording means for recording the physical quantity extracted from the average power spectrum extraction means 16, and 18 is a difference between the recorded content and the newly input physical quantity. 19 is a determination means for determining the degree of similarity between the physical quantity in the recording means 17 and the new large blade physical quantity based on the magnitude of the correction amount; 19 & is a determination means for pattern matching when the degree of similarity in the determination means 19 is high; A switch 19b that connects the means 12 and the determining means 13 is a switch 19b that connects the determining means 12 and the determining means 13 when the first password is input and the determining means 19 recognizes the first password because the degree of similarity is high.
A switch 20 is connected to the input terminal 21 and operates to prevent the audio analysis means 9 from being connected to the energy extraction means 15, etc., and has a level adjustment amplifier and a variable band characteristic amplifier. This is a transfer characteristic correction means that changes the transfer characteristic using each correction amount in order to increase the similarity when the degree of similarity with the frequency characteristic is low. Reference numeral 22 denotes a time processing means which has two or more types of passwords in the speech recognition device and invalidates the previous speech recognition result if the next speech recognition is not performed within a certain period of time from the end of the previous speech recognition.

The operation of this embodiment configured as above will be described below.

First, the level and frequency characteristics of the first password at the time of voice registration are extracted in advance by the energy extraction means 16 and the average power spectrum extraction means 16, and the recording means 17
Record it in

Next, using the first password during voice recognition, the level frequency characteristics are recorded by the average power spectrum extraction means 16 and the recording means 1.
7, the data is compared with the data in the recording means 17 by the correction amount calculation means 18 to determine the correction amount, and the judgment means 19
If the amount of correction is large, the analysis data of the voice analysis means 9 is input to the transfer characteristic correction means 2o installed in front of the voice analysis means 9. The transfer characteristic correction means 20 is controlled in the same manner.

Furthermore, if the amount of correction is small, the analysis data is input to the data normalization means 10.

Next, if the first password passes, the determination means 13 outputs an output signal to the time processing means 22, the tape recorder is stopped, and the second password is waited for. Also, only if the second or subsequent password is entered within a certain period of time,
The data normalization means 10 starts the subsequent recognition operation.
The time processing means 22 sets a timer. If the second password (or third password) is not recognized within this time, there is a high probability that the voice is from someone other than the owner.
A message requesting the first password is sent again using a voice synthesis output means or the like. Furthermore, since there is a high possibility that the primary password is noise other than the password, the tape recorder starts operating, sends out the owner's absence message, and when the message ends, starts recording the message from the outside. At that time, do not enter voice recognition mode.

That is, if the owner is the owner, there is no need to listen to the message, and the password can be recognized within the time it takes to output the message.

Furthermore, after recognizing several strange passwords (after confirming that the owner is the owner), the word for controlling the tape recorder is "rewind".

It is recognized by the utterance of ``fast forward j,'' etc., and is output from the device control signal generating means 14 to control the tape recorder.

3 to 5 are flowcharts of operation signals of this embodiment. From the beginning of FIG. 3, the telephone line is connected, and after the tape message is played back, an answering machine is recorded by an outsider. At this time, when the first password is inputted by the owner's voice, level and frequency correction is performed to determine the first password. At this time, if it is recognized as the first password, the tape recorder is stopped and a timer is started. Furthermore, it is determined whether or not it is the second password, and if it is not recognized within the timer time, the process returns to the beginning and plays the tape.

After the third password is recognized in the same way, the tape recorder control voice commands "fast forward" and "rewind" are recognized and the tape recorder is controlled. In the case of fast forwarding, after the tape has finished recording, that is, at the end of the message recorded by an outsider, or at the end of the message that starts the timer, the timer is set to
) L, determines whether to "record" or "rewind" and records the owner's voice when recording. Also, if there is no voice within the timer operation, the line will be disconnected. Furthermore, if the first password is input within the specified time and "erase" is input, the message is erased from the beginning in message 8 of the tape recorder.

This is controlled from the outside when the tape is filled with messages from the outside and no more messages can be recorded, thereby preventing triple malfunctions.

FIG. 6 is a hardware configuration diagram of this embodiment, in which a transfer characteristic correction means 20 is connected to a receiving terminal 2111 of a telephone line converted into four wires via a two-wire and four-wire hybrid transformer.
as programmable amplifier 23, frequency correction amplifier 2
4f: A recognition and tape recorder control unit 26 is provided, which is indicated by a broken line, and is connected to the tape recorder 26, the programmable amplifier 23, and the frequency correction amplifier 24. Further, an explanation of the operation of the tape recorder 26 is sent to the transmission terminal 2tb along with a tape reproduction output signal using the voice synthesis means 27, so that messages and the operation of the tape recorder 26 can be confirmed from a remote telephone. 28 is an amplifier for the speaker 29.

When the series of operations according to this embodiment is performed by checking the synthesized sound, it is as follows.

(A) In response to the input of the first password, "Please enter the second password" is sent out using a synthesized voice.

(B) The second password 1 and the third password are recognized in the same way. If the 1st to 3rd passwords cannot be recognized, please enter the 1st password again.
A) Start over.

(C) If up to the third password is recognized, the message "Hassword completed" will appear and "Fast forward" will appear.

The same voice is sent out from the synthesized sound every time an operation such as "rewind" is performed.

(D) When the tape message ends, a synthesized sound such as "message desu" or "pippi" is sent out to convey the message end status, and "recording" and "erasing" operations are performed within one hour of the timer.

At this time, if the owner utters the word "record" while recording, the synthesized sound "Rokuon" is sent out and the tape recorder operates in recording mode.

(E) When "erasing", first enter the "recording" mode and enter the first password within the time limit, then utter "shokyo" from the synthesized voice for the first time with the "erase" voice, and record the message on the tape recorder. to erase.

Similarly, the hierarchical structure using passwords in this embodiment,
The means used to control the tape recorder can be applied to the control of air conditioners in household equipment and remote monitoring of disaster prevention monitoring equipment.

A quick audio band signal such as a password whistle, a dial oscillation sound of a telephone, etc. may be registered, and furthermore, it may be registered in the standard pattern memory 11 in combination with a password.

Effects of the Invention As described above, according to the voice remote control device of the present invention, it is possible to control an answering machine or a garden device from a distance using, for example, a telephone line using an easily usable voice band signal such as voice. Furthermore, if a hierarchical structure based on passwords is used, it is possible to easily protect confidentiality and avoid trouble caused by others, which has a great effect.

[Brief explanation of the drawing]

FIG. 1 is a block diagram of a conventional voice recognition device, FIG. 2 is a block diagram of a voice remote control device according to an embodiment of the present invention, and FIG.
6 through 6 are flowcharts of the same operation signals, and FIG. 6 is a configuration diagram of the same hardware. 9...Speech analysis means, 10...Data normalization means, 11...Standard speech pattern memory, 12.
... Pattern matching means, 13 ... Judgment means, 14 ... Equipment control signal generation means, 15 ...
Energy extraction means, 16: Average power spectrum, extraction means, 17: Recording means, 18: Correction amount calculation means, 19: Judgment means, 2o: Transfer characteristic 4 corrected pitches. Name of agent: Patent attorney Toshio Nakao and one other person. 3rd page 4th page

Claims (1)

[Scope of Claims] (1) Speech analysis means for analyzing input speech band signal phrases, transfer characteristic correction means for inputting the correction amount determined by the transfer characteristic difference extraction means, and data determined by the voice analysis means. a standard speech pattern memory for registering data at the time of registration, a pattern matching means for matching the input speech pattern with the standard speech pattern in the standard speech pattern memory at the time of recognition, and matching of the pattern matching means. a voice recognition device having a determination means for determining the degree of similarity between voice inputs at the time of registration and at the time of recognition from the output;
An audio remote control device that controls equipment using the audio in-band signal. (2) The voice remote control device according to claim 1, which has a hierarchical structure of two or more types of voice passwords and controls the operation of the device after recognizing all the voice passwords. (3) The transfer characteristic difference extraction means includes an energy extraction means for extracting the maximum value or average energy of the registered voice at the time of voice registration in the first voice password processing, and an average power by frequency band of the count analyzed by the voice analysis means. Average power spectrum extraction means for extracting a spectrum; recording means for recording physical quantities extracted by the energy extraction means and the average power spectrum extraction means; energy extraction means and average power spectrum extraction means for input signals during speech recognition. and a correction amount calculation means for extracting a correction m from the physical quantity extracted in the recording means and the physical quantity in the recording means extracted at the time of voice registration, and the transfer characteristic correction means corrects the transfer function of the signal transmission system using the correction amount. An audio remote control device according to claim 1. (4) A patent claim in which the voice recognition device performs voice password processing using a time processing means to request the first voice password again if the next voice password is not recognized within a certain period of time after recognizing one voice password. The voice remote control device according to item 1. (6) The voice remote control device according to claim 2, wherein a password hierarchy structure is applied to the operation control with high importance among the device operation processing of the voice recognition device, and the operation command is recognized after passing the voice password. (6) A voice band signal other than the owner's voice pattern, such as a whistle, is included in the password registration pattern of the voice recognition device.
3. The audio remote control device according to claim 1, wherein the audio remote control device performs the recognition operation by registering audio band signals such as the sound of clicking one's lips or the dial oscillation sound of a telephone.