JPH08314493A - Voice recognition method, numeral line voice recognition device and video recorder system - Google Patents

Abstract

PURPOSE: To provide a voice recognition method and a numeral line voice recognition device improving handleability and obtaining a precise recognition result. CONSTITUTION: This method is the voice recognition method constituted of processing Pi inputting a voice signal from a microphone 11 as a recognition subject, the processing P2 obtaining a recognition score showing the probability of the recognition related to plural respective recognition candidates, the processing P31 performing first decision comparing the first rank recognition score with a first reference value and performing rejection when the first rank recognition score is smaller than the first reference value, the processing P32 obtaining a score ratio showing the ratio of a second rank recognition score for the first rank recognition score when the first rank recognition score is larger than the first reference value, and performing second decision comparing the score ratio with a second reference value and performing the rejection when the score ratio is larger than the second reference value and the processing P4 regarding the recognition candidate obtaining the first rank recognition score as the recognition result when the score ratio is smaller than the second reference value.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a voice recognition method and a number string voice recognition device, and more particularly to a voice recognition method and a number string voice recognition device suitable for voice recognition of a number string uttered by a speaker.

[0002]

2. Description of the Related Art It is said that it is difficult for people who are not good at machine operation, especially for elderly people, to make a recording reservation operation of a video tape recorder which is widely used as a video recorder. In recent years, a technique for recording reservation of a video tape recorder using a so-called number sequence has become widespread. This G code is generally such that when a digit string of about 2 to 8 digits is input to a dedicated device, the digit string is decoded and the data necessary for recording reservation of the video tape recorder, that is, the channel number and the recording. The start date and time, the recording time, etc. are set in the video tape recorder. In other words, the G code is generated by encoding the above-mentioned specifications according to a predetermined law.
The G code generated in this way is published in newspapers and the like. Therefore, the user only needs to enter the number string into the dedicated device while looking at the G code published in the newspaper or the like, so that even those who are not good at machine operation can easily make a recording reservation for the video tape recorder.

However, in reality, it is unexpectedly troublesome to sequentially input each number of an unknown number string by key operation, and it is often necessary to erroneously input the number again from the beginning. This is because, for example, if the user inputs a telephone number stored by himself / herself, the user can operate the key by putting his / her eyes on only the key, but when inputting an unknown number string such as a G code. This is because the user needs to perform a key operation while reciprocating his or her line of sight between the number string published in a newspaper or the like and the key of the device.

Under these circumstances, it is possible to input the G code by voice. In that case, the user only has to read each number in order with the line of sight fixed on the G-coded number string published in newspapers, etc., which may lead to a state where the user must re-enter the number on the way. Will be less.

On the other hand, in recent years, a wireless telephone called a so-called mobile telephone has become widespread. Mobile telephones are roughly divided into car telephones mainly installed in automobiles and mobile telephones carried by users, and there are also types that can be used for both. From the viewpoint of safety, it is desirable that the car phone be of a so-called hands-free type that can be operated by the driver without touching it. Hands-free type telephones have already been put to practical use even after being connected to a line even in ordinary telephones, but it is necessary for the user to operate the dials themselves. However, particularly in the case of a car telephone, when the user operates the dial buttons while driving, his / her eyes and his / her eyes are caught, which is very dangerous. Therefore, it is desirable that a telephone number can be input by voice only by first turning on the main switch, and after the line is connected, hands-free communication can be performed as it is.

By the way, it is difficult for the current speech recognition technology to speech-recognize all the words of everyone, and in general, only a limited range of words by the speech of a person who is registered in advance is targeted for recognition. In many cases But,
Even if different readings are included, it is possible to recognize everyone's voice by limiting the recognition target to only numbers expressed in about 15 different words. Therefore, the voice recognition device for everyone can be sufficiently put to practical use for the purpose of only the numbers such as the input device of the G code or the input device of the telephone number as described above.

FIG. 19 is a schematic diagram showing the concept of the principle of voice recognition by the conventional voice recognition device used for the above-mentioned applications. Further, FIG. 20 is a flowchart showing a specific processing procedure by the conventional voice recognition device shown in FIG.

In FIG. 19, reference numeral 1 indicates a microphone, and the voice uttered by the speaker is input. The voice input to the microphone 1 is captured as a voice signal by the voice input process P1 and recognized for each of a plurality of recognition candidates from "0 (zero, ray, circle)" to "9 (kyu, ku)". The recognition score indicating the accuracy of is obtained by the voice recognition process P2. Whether or not the recognition candidate for which the first-ranked recognition score has been obtained by this voice recognition processing P2 is valid is checked in the reject check processing P3 by comparing the reference value with the first-ranked recognition score. . Then, in the reject check process P3,
If the first-ranked recognition score is larger than the reference value, it is recognized as valid, and the recognition result notification process P4 utters a recognition candidate from the speaker 6 in a synthesized voice, or LC
Notify by displaying characters on D 7.

Next, a specific processing procedure when the conventional speech recognition method as described above is executed by the speech recognition apparatus will be described with reference to the flowchart of FIG. First, a reject threshold "reject", which is a reference for voice recognition, is set (step S1). This reject threshold "reje
The ct ”setting may be a fixed value or may be changed according to various conditions. Next, when an input button (not shown) of the device is turned on (step S2), the voice from the microphone 1 is output. Is input and captured as a voice signal by the voice input process P1 (step S3).

The voice signal taken in by the voice input process P1 is recognized by the voice recognition process P2 by a conventionally known method (step S4). Specifically, in the voice recognition process P2, “0” to “” are applied to the captured voice signal.
A recognition score representing the accuracy of the recognition result is calculated for each of the 9 ”recognition result candidates.

In the reject check process P3, the first-ranked recognition score of the recognition scores calculated by the voice recognition process P2 is first set as the reject threshold "re".
It is determined whether it is larger than "ject" (step S5).
If the result of this determination is NO, the rejection is informed that proper recognition could not be performed (step S10), and the voice signal is captured again by the voice recognition process P1.
(Step S3). If the determination result in step S5 is YES, the recognition result notification process P4 causes the speaker 6 or LC
Notification of recognition results using D 7 is performed (step
S6).

In this state, if the transfer button (not shown) provided in the apparatus is not operated (step S
7) Then, it is temporarily stored as the recognition of one number is completed (step S11). Then, the process returns to step S3 to accept the input of the next number. As described above, the steps S3, S4, S5, S6, S7, and S11 are sequentially executed, and the loop process of returning to step S3 is repeated, so that one numerical value for each loop process is repeated. Speech recognition is performed.

When the transfer button is turned on after all the numbers have been input in this way (step S
7) Then, the respective numbers input so far are decoded as a number string (step S8), and the decoding result is transmitted to a device such as a video tape recorder or a telephone which originally needs it. After this, the input button is turned off (step S9).

After the recognition result of one number is notified in step S6, as shown in step S12, the user recognizes the recognition result, that is, the recognition result uttered from the speaker 6. Synthetic voice or LCD
It is also possible to use a confirmation button to confirm the number of the recognition result displayed on 7.
This is because even if the recognition result candidate for which the first recognition score is obtained by the reject check unit 4 is a candidate for the recognition result, it is not necessarily guaranteed that it is a valid recognition result.

Now, for example, a case where the speaker inputs the numeral string "5492" by voice will be described with reference to the schematic diagrams of FIGS. 21 and 22 showing the state of the response of the conventional voice recognition apparatus to it. In the conventional voice recognition device, when the device performs voice recognition, a confirmation sound indicating that is emitted and the recognition result is displayed, and the user operates the confirmation button to input one number. Type to complete,
There is a type in which the recognition result is uttered by synthetic voice, and if it is correct, the user utters the next number as it is.

First, FIG. 21 shows the former type. That is, when the user utters “go”, the device utters a confirmation sound such as “beep” from the speaker 6 when it recognizes the voice, and the recognition result is displayed.
Display on LCD 7. The user shifts his line of sight on the LCD 7 to check the displayed recognition result, turns on the confirmation button, and then utters the next number, “Yon”.

In the example shown in the schematic diagram of FIG. 21, the input intervals of individual numbers are not shortened to a certain extent or more, and therefore, the intervals for uttering each number are long from the user's standpoint, which is convenient. Becomes worse. In addition, since it is necessary to shift the line of sight to the LCD 7 each time the confirmation sound is emitted, there is a problem in applying it to a mobile phone.

The latter type is shown in FIG. That is, first, when the user utters "go", the device utters the recognition result as a synthesized voice from the speaker 6 when recognizing the voice. After confirming the recognition result uttered by this synthetic voice, the user utters the next number, "Yong."

In the example shown in the schematic view of FIG. 22, the user can operate the LCD 7 in comparison with the example shown in FIG.
Although it is not necessary to shift the line of sight to the upper part, the input interval for each character is not shortened to a certain extent, and therefore, from the user's point of view, the interval for uttering each number is increased and the usability deteriorates.

[0020]

By the way, the conventional speech recognition apparatus as described above has the following two disadvantages.

The first problem is the accuracy of the recognition result. As described above, in the conventional method for general speech recognition, if the first-ranked recognition score of some recognition result candidates is equal to or higher than the reject threshold “reject”, the candidate is finalized. It was a recognition result. However, even if there is no significant difference between the recognition scores of two or more recognition result candidates, the candidate for which the first-ranked recognition score is obtained is conventionally a valid recognition result. , The accuracy of the recognition result was low. For example, if the speaker says "ichi", "shichi", or "hichi",
As a result of the recognition, there is a possibility that almost the same recognition scores can be obtained for both "1" and "7", which increases the possibility of erroneous recognition.

The second problem is that the conventional speech recognition method has a trade-off relationship between reducing erroneous recognition and reducing the occurrence of rejects. Generally, in order to reduce the possibility of erroneous recognition in voice recognition, the reject threshold may be set to a relatively high value. However, the higher the reject threshold is, the more rejects occur. Under such circumstances, the applicant of the present application has filed Japanese Patent Application Laid-Open No. 4-152397 (G10L 3/00 3
The invention as disclosed in (01) has been filed first.

In the speech recognition apparatus disclosed in Japanese Patent Laid-Open No. 4-152397, the reject threshold is set lower in the re-recognition process after being rejected once in the recognition process of a certain recognition target. Thus, by dynamically setting the reject threshold, the problem that the speaker is rejected no matter how many times the same voice is generated is solved. However, the above-mentioned first problem cannot be solved.

The third problem is the usability of the device. This is for the purpose of solving the above-mentioned first problem. When displaying the recognition result of each numeral on the LCD, for example, and asking the user for confirmation, the user moves the line of sight to the LCD and operates the confirmation button. Need to do. Such operation is not only troublesome for the user,
It is very dangerous when driving a car,
There is a time lag before the next input becomes possible.
Further, when uttering the recognition result of each number as a synthesized voice from the speaker and asking the user for confirmation, the user does not have to move the line of sight, but at least the confirmation button needs to be operated. Even in this case, since there is a time lag until the next input is possible, it cannot be denied that usability deteriorates.

The present invention has been made in view of the above circumstances, and improves the usability, obtains a more accurate recognition result, and avoids a large number of rejects. An object is to provide a recognition device and a video recorder system.

[0026]

A speech recognition method according to the present invention obtains a recognition score indicating the accuracy of a recognition result for a speech signal to be recognized for each of a plurality of recognition candidates,
The first recognition score of the obtained recognition scores is the first
The first judgment is made by comparing with the reference value of, and as a result of the first judgment, if the recognition score of the first place is smaller than the first reference value, all recognition candidates are rejected and the first place If the recognition score is higher than the first reference value, a score ratio representing the ratio of the second-ranked recognition score to the first-ranked recognition score is obtained, and the obtained score ratio is compared with the second reference value. If the score ratio is larger than the second reference value as a result of the second judgment, all recognition candidates are rejected, and the score ratio is smaller than the second reference value. In this case, the recognition candidate for which the first recognition score is obtained is regarded as a recognition result.

Further, in the voice recognition method according to the present invention, when the same recognition target is rejected according to the result of the first judgment or the second judgment, the first reference value is made smaller. It is characterized in that the second reference value is set to a larger value and the first determination and the second determination are performed again.

Further, in the voice recognition method according to the present invention, the first reference value is obtained when the rejects according to the result of the first judgment or the second judgment for the same recognition object are continuously performed a predetermined number of times. Is set to "0" and the second reference value is set to "1".

The number string voice recognition apparatus according to the present invention calculates a recognition score indicating the accuracy of a recognition result for a voice signal of each one-digit number of a series of number strings to be recognized, for each of a plurality of recognition candidates. The calculation means and the first-ranked recognition score calculated by the recognition score calculation means are the first
And a first determining means for comparing with the reference value of No. 1 and a first determining means rejects all recognition candidates when the first-ranked recognition score is smaller than the first reference value.
And a score representing the ratio of the second-ranked recognition score to the first-ranked recognition score calculated by the recognition-scoring calculation means when the first-ranked recognition score is larger than the first reference value. The score ratio calculating means for obtaining the ratio, the second judging means for comparing the score ratio calculated by the score ratio calculating means with the second reference value, and the score ratio as the result of the judgment by the second judging means are the second. If the score ratio is smaller than the second reference value, the first recognition score is determined by the second reject means that rejects all the recognition candidates if the recognition score is greater than the reference value. And a recognition result output means for outputting the calculated recognition candidate as a recognition result.

The number string voice recognition apparatus according to the present invention is
When the same reject target is rejected by the first reject means or the second reject means, the first reference value is set smaller and the second reference value is set larger. It is characterized in that the first judging means and the second judging means are provided with re-judging means for making the judgment again.

Further, the number string voice recognition device according to the present invention is
The re-determination means sets the first reference value to "when the rejection by the first reject means or the second reject means for the same recognition target is continuously performed a predetermined number of times.
It is characterized in that the second reference value is set to 0 "and the second reference value is set to" 1 ".

Furthermore, the number string voice recognition apparatus according to the present invention causes the voice output means and the voice output means to utter a predetermined synthetic voice each time the recognition result output means outputs a recognition result, and a series of recognition targets is output. When the recognition result output unit outputs the recognition result for all the numbers in the number string, the voice output unit further includes a unit for continuously uttering each number of the recognition result as a synthetic voice. To do.

Further, the video recorder system of the present invention comprises a video recorder main body which is designed to make a recording reservation by being given data of a predetermined format, and a remote control device which transmits data of a predetermined format. The number string voice recognition device of the present invention as described above is used as a remote control device, and a voice signal of each one-digit number of a series of number strings which is the source of data to be given to the video recorder main body is generated. It is a recognition target.

[0034]

In the voice recognition method according to the present invention, the recognition score indicating the accuracy of the recognition result for the voice signal to be recognized is obtained for each of the plurality of recognition candidates, and the first recognition score among the obtained recognition scores is obtained. Is compared with a first reference value, and if the result of the first determination is that the first-ranked recognition score is smaller than the first reference value, all recognition candidates are rejected. , If the first-ranked recognition score is higher than the first reference value, a score ratio representing the ratio of the second-ranked recognition score to the first-ranked recognition score is calculated, and the calculated score ratio is A second determination is made to compare with a reference value of 2, and if the result of this second determination is that the score ratio is larger than the second reference value, all recognition candidates are rejected and the score ratio is set to the second. If it is smaller than the reference value of Core recognition candidates obtained is regarded as a recognition result.

Further, in the voice recognition method according to the present invention, the first reference value becomes smaller when the same recognition target is rejected according to the result of the first judgment or the second judgment. The first determination and the second determination are performed again after the second reference value is set to a larger value.

Further, in the voice recognition method according to the present invention, the first reference value is obtained when the rejects according to the result of the first judgment or the second judgment for the same recognition object are continuously performed a predetermined number of times. Is set to "0" and the second reference value is set to "1".

In the number string voice recognition apparatus according to the present invention, the recognition score calculation means for each of the plurality of recognition candidates has the recognition score indicating the accuracy of the recognition result for the voice signal of each one-digit number of the series of number strings to be recognized. The first-ranked recognition score calculated by the recognition score calculation means is compared with the first reference value by the first judgment means, and the result of the judgment by the first judgment means is the first-rank recognition. If the score is smaller than the first reference value, all recognition candidates are first
When the recognition score of the first place is higher than the first reference value, the ratio of the recognition score of the second place to the recognition score of the second place calculated by the recognition score calculation unit is The score ratio representing means is obtained by the score ratio calculating means, the score ratio calculated by the score ratio calculating means is compared with the second reference value by the second judging means, and the result of the judgment by the second judging means is the score ratio. Is larger than the second reference value, all recognition candidates are rejected by the second rejecting means, and the score ratio is the second.
When it is smaller than the reference value of, the recognition candidate for which the first recognition score is calculated by the recognition score calculation means is output as the recognition result by the recognition result output means.

Further, in the numeral string voice recognition apparatus according to the present invention, when the first reject means or the second reject means rejects the same recognition target, the re-determining means determines the first reference value. The second reference value is set to a smaller value, and then the determination is performed again by the first determining means and the second determining means.

Further, in the number string voice recognition apparatus according to the present invention, when the first reject means or the second reject means rejects the same recognition object continuously for a predetermined number of times, the re-determining means determines The reference value of 1 is set to "0" and the second reference value is set to "1".

Furthermore, in the numeral string voice recognition apparatus according to the present invention, each time the recognition result output means outputs the recognition result, a predetermined synthetic voice is uttered from the voice output means, and all the series of numeral strings to be recognized are recognized. When the recognition result output means outputs a recognition result for a numeral, each numeral of the recognition result is continuously uttered by the voice output means as a synthetic voice.

Further, in the video recorder system of the present invention, the voice signal of each one-digit number of the series of numbers which is the source of the data to be given to the video recorder main body is recognized as the number of the present invention as described above. Voice recognition is performed by the remote control device using the row voice recognition device.

[0042]

DESCRIPTION OF THE PREFERRED EMBODIMENTS The present invention will be described below in detail with reference to the drawings showing the embodiments thereof. In each of the embodiments described below, the case where the voice recognition method and the number string voice recognition device of the present invention are applied to a G code input device (hereinafter referred to as a G code remote controller) of a video recorder system will be described. In each of the following embodiments, the device configuration, that is, the hardware is common, but the software (recognition method) in each embodiment is different. Therefore, first, the principle of the voice recognition method according to the present invention will be described, and then the hardware configuration common to each embodiment will be described.

FIG. 1 is a schematic diagram showing the principle of the voice recognition method according to the present invention. In FIG. 1, reference numeral 11 indicates a microphone into which a voice uttered by a speaker is input. Voice input processing for voice input to microphone 11
A speech recognition process P2 obtains a recognition score indicating the accuracy of recognition for each of the recognition candidates "0 (zero, ray, circle)" to "9 (kyu, ku)" that is captured as a voice signal by P1. This voice recognition process P2
Whether the recognition candidate for which the rank recognition score is obtained is valid or not is determined by comparing the first reference value with the recognition score of the first rank, thereby performing a reject check process P31 based on an absolute difference.
Checked in.

In the reject check process P31,
If the first-ranked recognition score is smaller than the first reference value, all recognition candidates are rejected. If not rejected, further reject check processing by relative ratio
In P32, the ratio of the second-ranked recognition score to the first-ranked recognition score is determined, and the ratio is compared with the second reference value to perform a rejection check based on the relative ratio. Then, both reject check processes P31, P32
If both are recognized as valid, the recognition candidate of the first recognition score is regarded as the recognition result, and the recognition result notifying process P4 causes the speaker 15 to make a synthetic voice or the LCD 25 It is notified by being displayed in characters.

FIG. 2 is a block diagram showing a hardware configuration example of a G code remote controller of a video recorder system as a specific application example of the numeral string voice recognition apparatus according to the present invention. The internal structure of the number string voice recognition device of the present invention is roughly represented by reference numeral 10 (DSP (Digital Signal P
rocessor) section and a microcomputer section indicated by reference numeral 20. Reference numeral 40 is a video recorder main body which constitutes a video recorder system together with the G code remote controller. In this embodiment, a video tape recorder (VTR) is used.

The DSP section 10 is mainly composed of the microphone 11, A /
D converter 12, DSP13, D / A converter 14, speaker 15, PROM16,
I / F (interface) with the microcomputer unit 20,
It consists of RAM18, ROM19, etc.

The voice uttered by the speaker is input from the microphone 11, taken in as a digital signal by the A / D converter 12 via the microphone amplifier 11A, and processed by the DSP 13. Storage devices such as PROM16, RAM18, and ROM19 are connected to the DSP13.

The ROM 19 stores in advance coded data of various voices to be output as synthesized voices from the speaker 15, weight data of a neural network for voice recognition, and the like. The RAM 18 temporarily stores the values of various variables necessary for processing by the DSP 13, data during processing, processing result data, and the like. The PROM16 stores the programs required for processing by the DSP13.

The DSP 13 performs a predetermined process according to the program stored in the PROM 16 and, if necessary, the ROM 19
Read the encoded data stored in
By outputting to the speaker amplifier 15A via the / A converter 14, a synthesized voice is uttered from the speaker 15. Also,
The DSP 13 sends and receives data to and from the microcomputer 21 as needed. I / F 17 is used for this.

The microcomputer section 20 mainly includes a microcomputer 21, a transfer button 22, and a G code button.
23, I / F24, LCD25, transmitter 26, etc.

The transfer button 2 is provided on the microcomputer 21.
2, The ON signal is continuously given from the G code button 23 when they are turned on (pushed). Transmitter
26 is used for transmitting the result of decoding the recognized digit sequence as a G code to the VTR 40. Also, LCD
25 is used by the microcomputer 21 via the I / F 24 to display a numeric string of the recognition result or other various displays.

ROM and RAM are provided in the microcomputer 21.
Is built-in, and the built-in ROM has a microcomputer
A program necessary for the processing by 21 or a program (or table) for decoding an arbitrary number string as a G code is stored. Further, the built-in RAM stores various data while the program is being executed by the microcomputer 21.

FIG. 3 is a schematic view showing the external appearance of a G code remote controller as a number string voice recognition device of the present invention. In FIG. 3, reference numeral 30 is a casing in which the above-mentioned components shown in FIG. 2 are stored. The microphone 11, the speaker 15, the transfer button 22, the G code button 23 and the LCD 25 are partially exposed on the surface of the casing 30 in order to effectively exhibit their respective functions.

[First Embodiment] The operation of the first embodiment of the G code remote controller as the numeral string voice recognition device of the present invention having the above-described configuration, in other words, the voice recognition method of the present invention, Description will be made with reference to the flowcharts of FIG. 4, FIG. 5, and FIG. The flowcharts shown in FIGS. 4, 5 and 6 will be referred to as a first flowchart hereinafter.

The feature of the first flow chart shown in FIGS. 4, 5 and 6, that is, the main feature of the present invention is that the recognition of a voice-input numeral is simply performed by the first method.
First, instead of using only the absolute value of the recognition score
The point is that the ratio between the recognition score of the second rank and the recognition score of the second rank is also taken into consideration. Due to such characteristics, the first embodiment of the present invention prevents frequent occurrence of erroneous recognition.

First, the microcomputer 21 sets the variable j,
Clears both k to zero (step S31), and G code button
Check the on / off state of 23 (step S32). When the G code button 23 is in the ON state, the microcomputer 21 clears the timer value "time" to zero (step S33). Next, the microcomputer 21 sends a voice capture command to the DSP 13 via the I / F 17 (step S34). In response, the DSP13 has a microphone 11
The voice input from will be accepted.

When the DSP 13 captures the voice, it sends a voice capture end response to the microcomputer 21 via the I / F 17. When the microcomputer 21 receives the voice capture end response transmitted from the DSP 13 (step S35), the microcomputer 21 transmits a command (recognition command) for causing the DSP 13 to perform voice recognition (step S36). However, when the DSP 13 cannot take in the sound and the G code button 23 is turned on (step S61), the loop process that goes back and forth between steps S35 and S61 is repeated, and the DSP 13 causes the microphone 11 to operate. Waiting for voice input from.

In response to the recognition command transmitted from the microcomputer 21, the DSP 13 recognizes the voice previously captured, and transmits the result, that is, the candidate of the recognition result and the recognition score of each candidate to the microcomputer 21. To do. The microcomputer 21 receives the recognition result candidates and the recognition score of each candidate transmitted from the DSP 13 (step S37). Note that the procedure of voice recognition by the DSP 13 at this time uses known technology.

Next, the microcomputer 21 sets "150" as the reject threshold "reject" and the score ratio threshold "
"0.5" is set as each score (step S38). These values may be fixed by the program stored in the internal ROM or may be changeable according to various conditions. Then, the microcomputer 21 compares the first-ranked recognition score among the recognition scores transmitted from the DSP 13 with the reject threshold “reject (= 150)” (step S39). When the recognition score is equal to or higher than the reject threshold “reject”, the microcomputer 21 performs the following calculation (step S4
0). In this example, the theoretical minimum value of the recognition score is "
0 ", the maximum value is" 255 ".

Score ratio = second recognition score / first recognition score

The microcomputer 21 executes this step S4
The score ratio obtained in 0 and “0.5” in step S38 previously.
Compare with the score ratio threshold "score" set to "
(Step S41). As a result, the score ratio is the score ratio threshold "s.
If it is equal to or more than "core", the microcomputer 21 stores the recognition candidate having the first-ranked recognition score in the built-in RAM as the recognition result and transmits the response voice output command to the DSP 13 (step S42). In response to the response voice output command, the DSP 13 outputs the number of the recognition result as a synthesized voice from the speaker 15 and outputs an end response to the microcomputer 21.

Upon receipt of the end response from the DSP 13 (step S43), the microcomputer 21 increments the variable j by "1" (step S44) and sets "0" in the transfer flag (step S45).

With the above, one number is recognized as voice, and the process is returned to step S32. The variable j is incremented by "1" each time the above-mentioned loop processing from step S32 to step S45 is executed once. Therefore, the value of the variable j represents the number of characters already recognized at that time.

By the way, when the first-ranked recognition score is less than the reject threshold "reject" in step S39, in other words, when the first-ranked recognition score is less than 150, and in step S41, When the ratio is less than the score ratio threshold “score”, the microcomputer 21 sends a command to the DSP 13 to output a reject sound indicating that a valid recognition result was not obtained, that is, the reject was performed. (Step S46). In response to this, the DSP 13 outputs a reject sound such as "buzz" from the speaker 15, and outputs an end response to the microcomputer 21. This DSP13
When the end response output from is received (step S47),
The microcomputer 21 returns the process to step S32.

When the microcomputer 21 detects that the G code button 23 is turned off in a state where the voice is not captured by the DSP 13 in the loop processing in which the steps S35 and S61 are repeated, the microcomputer 21 Sends a voice capture forced termination command to DSP 13 (step S62). In response to this, DSP13
Sends an end response signal, and upon receiving it, the microcomputer 21 copies the value of the variable j representing the number of characters recognized so far to the variable k and stores it.
(Step S64). After that, the microcomputer 21 clears the variable j to zero (step S65) and returns the process to step S32.

The above-described state is a state in which the user stops inputting any more digits after the number string having an arbitrary number of digits is input, in other words, whether or not the user has finished inputting the desired number string,
Alternatively, the user has stopped inputting numbers for some reason. In this case, the process returns to step S32, but if the G code button 23 is in the off state at that time, the process proceeds to step S71. This step
In S71, the microcomputer 21 checks the on / off state of the transfer button 22. If the transfer button 22 is not turned on, it is checked whether or not the time value "time" of the timer, which has been zero-cleared in step S33, has reached a predetermined time limit (step S78).

Until the timed value "time" of the timer reaches the time limit, the processing is returned from step S78 to step S32, and the loop processing through steps S71 and S78 is repeated. Therefore, when both the G code button 23 and the transfer button 22 are off, the microcomputer 21
Enters a state waiting for one of them to turn on. However, if all of the timers are in the off state and the timed value "time" of the timer reaches the time limit, the microcomputer 21 ends the loop processing described above, and thus the recognized number string is invalid. . However, if the G code button 23 is turned on before that, the microcomputer 21 detects it in step S32 and advances the process to step S33, so that a new digit string can be input. On the other hand, if the transfer button 22 is turned on, the microcomputer 21 detects it in step S71 and outputs the G code.
Performs processing for transmission to VTR40.

In this case, the microcomputer 21 first clears the timer value "time" to zero (step S7).
2), it is checked whether or not the value of the variable k is "0" (step S73). As described above, the variable k represents the number of recognized numbers. Therefore, when the value of the variable k is "0", the numeric string is not recognized, and the microcomputer 21 returns the process to step S32. Variable k
If the value of is not "0", then the microcomputer 21
Checks the value of the transfer flag (step S74).

When the value of the transfer flag is "0", it means that the transfer of the recognized digit string at that time has not been performed yet, and when it is "1", the recognized character string at that time. Each indicates that a column transfer has already occurred. Therefore, the microcomputer 21 returns the processing to step S32 if the value of the transfer flag is not "0",
If it is "0", the already recognized numeral string is decoded as a G code (step S75). The decoding result is transmitted to the VTR 40 by the transmitter 26 (step S76).
After this, the transfer flag is set to "1" (step S7
7), the process is returned to step S32.

As described above, in the first embodiment of the G code remote controller as the number string voice recognition device of the present invention, the recognition of the number inputted by the voice is simply performed by the absolute value of the first recognition score. A feature of the present invention is that it prevents occurrence of erroneous recognition by not only using the values but also considering the ratio between the first-ranked recognition score and the second-ranked recognition score. This will be specifically described below.

In the first flowchart shown in FIGS. 4, 5 and 6, the reject threshold "reject" is first set to 150 and the score ratio threshold "sc" is set in step S38.
ore "is set to 0.5, and the first-ranked recognition score is compared with the reject threshold" reject "in step S39. This process is a recognition generally used in the conventional speech recognition apparatus. However, it is assumed that a distribution of recognition scores as shown in the graph of FIG. 7 is obtained when the speaker utters “d”.

Here, the recognition score of the recognition result "Sun" is "183", and the recognition score of "D" is "1st".
It is 140 ”and it is in the second place. In the conventional case, the recognition score“ 183 ”of the first recognition result“ Sun ”is larger than the reject threshold“ reject (150 in this case) ”. Therefore, "Sun" is immediately regarded as a recognition result, resulting in a false recognition. However, in the first embodiment of the present invention, the ratio between the first-ranked recognition score and the second-ranked recognition score is obtained in step S40 of the first flowchart, and the score ratio threshold "score (this score If 0.
5) ”. In this case, the ratio between the recognition score“ 183 ”of the first recognition result“ San ”and the recognition score“ 140 ”of the second recognition result“ Dii ”is The recognition result is rejected because it is about 0.765 and is higher than the score ratio threshold “score”. In other words, even if the ambiguous recognition result is obtained such that there is no significant difference between at least the first-ranked recognition score and the second-ranked recognition score,
Even if the position recognition score is equal to or higher than the reject threshold “reject”, rejecting the recognition result prevents frequent misrecognition.

In the first embodiment of the G code remote controller as the numeral string voice recognition device of the present invention, a microcomputer is used.
Although the program 21 is programmed to perform the above-described processing, the procedure when the user actually uses the G code remote controller as the numeral string voice recognition device of the present invention will be described below.

Before the user starts using the G code remote controller as the number string voice recognition device of the present invention, the microcomputer 21 is in a state of waiting for some operation by the user. That is, since both the transfer button 22 and the G code button 23 are off,
The process proceeds in the order of S31, S32, S71, S78, and then steps
Returning to S32, the loop processing is repeated in the order of steps S71, S78 and S32.

When the user keeps the G code button 23 in the ON state to input the G code, an ON signal is given to the microcomputer 21. Microcomputer 21
Detects an ON signal from the G code button 23 in step S32 during the loop processing of steps S32, S71, S78, S32. As a result, the microcomputer 21 thereafter proceeds from step S32 to step S35, and enters a loop process that reciprocates between step S35 and step S61. When the user utters the first digit of the G code in this state, the DSP 13 captures the user's voice. After that, the microcomputer 21 executes the processing of each step from step S35 to step S45.
Recognizes the voice previously captured and utters the speaker 15 with the synthesized voice of the recognition result.

As described above, the microcomputer 21 repeats the loop processing of sequentially returning from step S32 to step S45 and returning to step S32, whereby the G code of an arbitrary number of digits uttered by the user sequentially. Each number of is taken one by one for each loop processing. And each captured number is the reject threshold ”
Speech recognition is performed with each of the reference values of "reject" = 150 and score ratio threshold value "score" = 0.5, and the synthesized speech of the recognition result is used to utter from the speaker 15.

It should be noted that if the recognition processing of the individual numbers is rejected midway, step S39 or step S41 is performed.
From step S46 to S47, the process proceeds to step S32
The process is returned to. Therefore, in this case, as long as the user continues to turn on the G code button 23, the rejected numeral can be input again.

After finishing inputting all the numbers of the G code, the user turns off the G code button 23. In this case, the processing by the microcomputer 21 proceeds from step S32 to step S3 earlier than the reaction time when the user confirms the response by the synthesized voice and turns off the G code button 23.
The user can select G
The fact that the code button 23 has been turned off is detected in step S61 during the loop processing in which steps S35 and S61 are repeated. As a result, the microcomputer 21 performs the processing of each step from step S62 to step S65 to end the voice acquisition in the DSP 13, and then advances the processing from step S32 to step S71.

In step S71, it is checked whether the transfer button 22 is on or off, and if it is off, it is checked in step S78 whether a predetermined time has elapsed. Further, step S7
When it is determined in 8 that the predetermined time has not elapsed, the process is returned to step S32, and thereafter, the loop process of sequentially returning to step S32 through steps S71 and S78 is repeated. Therefore, if the user turns on the transfer button 22 before the predetermined time elapses from the time when the last digit is recognized, the microcomputer 21 detects it at step S71 in the above-mentioned loop processing. In this case, the microcomputer 21 thereafter performs the processing of steps S72 to S77 to decode the previously recognized numeral string as a G code and decode it in the VTR.
Send to 40.

If the predetermined time elapses before neither the G code button 23 nor the transfer button 22 is turned on, step S7
Since it is detected at 8, the processing by the microcomputer 21 ends. In this case, the already recognized G code becomes invalid, but by turning on the G code button 23 again, the user can input a new G code.

If the user notices an erroneous input or erroneous recognition during the input of the G code, the G code button 23 may be turned off. In this case, the microcomputer 21 confirms that the G code button 23 has been turned off in step S6.
Since the detection is performed in step 1, the process is returned to step S32 as described above, and the loop process of returning from steps S71 and S72 to step S32 is repeated. Then, if the user turns on the G code button 23 again and keeps that state, the process proceeds from step S32 to step S33 of the above-described loop process, and a state in which a new G code can be input is entered.

In the first embodiment, the reject threshold "reject" is set to "150" and the score ratio threshold "score" is set.
Is set to "0.5" for voice recognition, but it goes without saying that other suitable numerical values may be adopted.

[Second Embodiment] Next, regarding the second embodiment of the G code remote controller as the numeral string voice recognition device of the present invention, the processing procedure of the microcomputer 21, in other words, the voice recognition of the present invention. 8 and 9 showing the method divided
Also, description will be made with reference to the flowchart of FIG. In addition,
Hereinafter, the flowcharts shown in FIGS. 8, 9 and 10 will be referred to as a second flowchart. In addition, FIG.
A first showing the processing procedure of the first embodiment among the processing steps of the second flowchart shown in FIGS. 9 and 10.
The same step numbers are given to the same portions as the processing steps in the flowchart of FIG.

In the second flow chart shown in FIGS. 8, 9 and 10, step S31 in the first flow chart shown in FIGS. 4, 5 and 6 is replaced by step S31. S131 replaces step S38 with step S1
38, step S144 instead of step S44
Step S165 is executed instead of S65, and step S48 is additionally executed after step S47 of the first flowchart shown in FIGS. 4, 5 and 6.

The characteristic of the second flow chart shown in FIGS. 8, 9 and 10, that is, the characteristic of the second embodiment is that when the recognition result of a numeral input by voice is rejected, Rejection threshold "reject" for input
And the score ratio threshold "score" is relaxed. In other words, in the second embodiment, the possibility of being rejected at the time of re-inputting after being once rejected is lower than that at the previous time, so that a large number of rejects are prevented.

The characteristics of the second embodiment as described above are mainly realized by the process of step S138 executed in place of step S38 of the first embodiment. That is, the microcomputer 21 first clears the variable i, which indicates the number of times the reject is consecutive, together with the other variables j and k to zero in step S131, starts the processing in step S32 and thereafter, and loops in the order of step S32, S71, S78. Start processing. Then, in the recognition processing in step S138 after capturing a certain number of voices, the variable i is "
Since it is 0 ", similar to the processing in step S38 of the first embodiment, the rejection threshold" reject "= 150 and the score ratio threshold" score "= 0.5 are set, and then step S39, Whether or not to reject is determined in S40 and S41.

If not rejected in the first recognition process, the variable j is incremented by "1" in step S144 instead of step S44, and the variable i representing the number of times the rejects are continuously performed is cleared to zero. However, if rejected, the variable i is incremented by "1" in step S48 following step S47.

As described above, when rejected in the first recognition process, the variable i is incremented by "1" and changed from "0" to "1". Therefore, in the next step S138, the reject threshold "reject" = 100. And the score ratio threshold "score" = 0.8, respectively.
In S39, S40, S41, it is determined whether or not to reject. Therefore, in the second recognition process after the first rejection, it is determined whether or not to reject in a state in which the possibility of rejection is smaller than that in the first recognition process.

Further, in the case of being rejected twice in succession, the variable i is further incremented by "1" in step S48 to become "2". Therefore, in step S138, the rejection threshold value "reject" = 0 and the score is again scored. Ratio threshold ”
score ”= 1.0, and then step S39,
Whether or not to reject is determined in S40 and S41. Therefore, at the time of the third recognition process after the second successive rejection, it is determined whether or not to reject in the state where there is no possibility of rejection. In other words, the recognition candidate of the first-ranked recognition score is unconditionally determined as the recognition result after being rejected twice in succession.

When the user finishes inputting the voice of the number string or cancels it in the middle, the variable i is cleared to zero together with the variable j in step S165 instead of step S65 shown in the first flow chart. After that, the process is returned to step S32.

The other processing is the same as that of the first embodiment described above, and the description thereof is omitted.

As described above, in the second embodiment of the G code remote controller as the numeral string voice recognition apparatus of the present invention, the recognition of the numeral inputted by voice is fixed to the fixed reject threshold as in the first embodiment. "Reject" and score ratio threshold "score
“Reject threshold“ reject ”, score ratio threshold” when re-input after rejected
It is characterized in that both the score and the score are relaxed. Due to such a feature, in the second embodiment of the present invention, in addition to the prevention of frequent occurrence of erroneous recognition in the first embodiment, the rejection It prevents frequent occurrence, which will be described in detail below.

In the second flow chart shown in FIG. 8, FIG. 9 and FIG. 10 described above, when recognizing a certain voice captured by the DSP 13, the first time rejecting is performed in step S138 as in the first embodiment. The threshold "reject" is 15
The score ratio threshold "score" is set to 0.5 and the first recognition score is compared with the reject threshold "reject" in step S39. here,
For example, assume that a distribution of recognition scores as shown in the graph of FIG. 11 is obtained.

Here, the recognition score of the recognition result "d" is "145", which is first, and the recognition score of "sun" is "90", which is second. In this case, the first
Since the recognition score “145” of the rank recognition result “d” is smaller than the reject threshold “reject (150 in this case)”,
Rejected. However, in the second embodiment of the present invention, if the user does not turn off the G code button 23, the input is accepted again, and in this step S138, the reject threshold "reject" is set to 100 and the score ratio threshold "
score ”is reset to 0.8 and then step
In S39, the first recognition score is again compared with the reject threshold "reject (100 in this case)".

In this case, the first recognition result is "d".
Since the recognition score “145” is larger than the reject threshold “reject”, the score ratio threshold “step” is determined in step S41.
score (0.8 in this case) ”. In this case, the first
Second, the recognition score of “D”, which is the recognition result of rank “145”
Since the ratio of the position recognition result “San” to the recognition score “90” is about 0.621, which is smaller than the score ratio threshold “score”, “D” is determined as the recognition result. In other words, in the second recognition process after being rejected once, the reject threshold “reject” and the score ratio threshold “sc
Since both ore "are relaxed, the possibility of being rejected is reduced. Furthermore, when rejected at the time of the second recognition processing, the reject threshold" reject "is" 0 ".
In addition, since the score ratio threshold “score” is set to “1” respectively, the rejection is not actually performed, and the candidate with the first recognition score is always determined as it is as the recognition result.

In the second embodiment of the G code remote controller as the numeral string voice recognition device of the present invention, a microcomputer is used.
21 is programmed to perform the above-mentioned processing, but the procedure for the user to actually use the second embodiment of the G code remote controller as the numeral string voice recognition device of the present invention will be described below. To do.

Before the user starts using the G code remote controller as the number string voice recognition device of the present invention, the microcomputer 21 is in a state of waiting for some operation by the user. That is, since both the transfer button 22 and the G code button 23 are off,
The process proceeds in the order of S131, S32, S71, S78, and thereafter, the process returns to step S32 and the loop process is repeated in the order of steps S71, S78, S32.

When the user keeps the G code button 23 in the ON state to input the G code, the ON signal is given to the microcomputer 21. Microcomputer 21
Detects an ON signal from the G code button 23 in step S32 during the loop processing of steps S32, S71, S78, S32. As a result, the microcomputer 21 thereafter proceeds from step S32 to step S35, and enters a loop process that reciprocates between step S35 and step S61. When the user utters the first digit of the G code in this state, the DSP 13 captures the user's voice. After that, the microcomputer 21 executes the processing of each step from step S35 to step S45.
Recognizes the voice previously captured and utters the speaker 15 with the synthesized voice of the recognition result.

As described above, the microcomputer 21 repeats the loop process of sequentially returning to step S32 through the steps from step S35 to step S45, whereby the G code of an arbitrary number of digits uttered by the user sequentially. For each loop processing, each number is captured one by one and the reject threshold is “reject” = 150, score ratio threshold ”
Speech is recognized with a reference value of "score" = 0.5, and the synthesized speech of the recognition result is uttered from the speaker 15.

If the input processing of individual numbers is rejected midway, step S39 or step S41 is performed.
From step S46 to S47, a reject sound such as "buzz" is uttered, the variable i is incremented by "1" in step S48, and then step S3.
The process is returned to 2. Therefore, in this case, if the user continues to turn on the G code button 23, the variable i becomes "1", and thus the reject threshold "reject" = 100 and the score ratio threshold "score" = 0.8, which is more than the first time. Re-recognition is possible when the reject reference value is low. But,
When rejected twice in succession, the variable i is "
When it becomes 2 ”, the rejection threshold“ reject ”= 0 and the score ratio threshold“ score ”= 1.0, and the recognition candidate of the first recognition score is unconditionally determined as the recognition result. Therefore, the rejection is repeated. The possibility that the input voice is reduced is recognized, and at the maximum, the third input voice after two consecutive rejects is recognized.

When all the numbers of the G code have been input, the user turns off the G code button 23 and then the transfer button.
Turn on 22. The subsequent operations are the same as those in the first embodiment described above.

In the second embodiment described above, the reject threshold "reject" is set to "150" every time reject is performed.
"," 100 "," 0 ", score ratio threshold" score "
Are sequentially changed to "0.5", "0.8", "1" in 3 steps, but may be changed in 2 steps or 4 steps or more. It goes without saying that numerical values may be adopted.

[Third Embodiment] Next, regarding the third embodiment of the G code remote controller as the numeral string voice recognition device of the present invention, the processing procedure of the microcomputer 21, in other words, the voice recognition of the present invention. Figures 12 and 13 showing the method divided
Also, description will be made with reference to the flowchart of FIG. In addition,
Hereinafter, the flowcharts shown in FIGS. 12, 13 and 14 will be referred to as a third flowchart. Also, in FIG.
A first showing the processing procedure of the first embodiment among the processing steps of the third flowchart shown in FIGS. 13 and 14.
The same step numbers are given to the same portions as the processing steps in the flowchart of FIG.

In the third flow chart shown in FIGS. 12, 13 and 14, step S42 in the first flow chart shown in FIG. 4, FIG. 5 and FIG. 6 is replaced with step S42. S142 replaces step S65 with step S2
65 to step S270 are executed respectively.

The characteristic of the second flowchart shown in FIGS. 12, 13 and 14, that is, the characteristic of the third embodiment, is that the user confirms the recognition result of each digit input by voice each time. Although it is not possible, it is possible to shorten the input interval of each number by configuring so that the whole can be confirmed when the input of the number string is completed.

The features of the third embodiment as described above are mainly realized by the processes of steps S265 to S270. That is, if the recognition result is not rejected during the process of capturing a voice of one number, the microcomputer 21 causes the DSP 13 to perform the process in step S142.
By sending a command for outputting a recognized confirmation sound to the user, for example, a sound such as a beep, the user is notified only that one number has been input and recognized.

When the voice input of the G code is completed or is stopped midway, the recognized digit is obtained by the processing of steps S265 to S270 instead of step S65 shown in the first flowchart. The rows are collectively uttered from the speaker 15 as synthetic voice. That is, the microcomputer 21 first determines whether the value of the variable k, that is, the number of recognized numbers is 1 or more, in other words, whether or not even one number is recognized (step S2
65).

If it is determined in step S265 that no numeral has been recognized, the microcomputer 21 returns to step S32 and enters a standby state.
If any number is recognized in step S265, the microcomputer 21 outputs a synthesized voice corresponding to the recognition result of the "kj" th recognized voice to the speaker.
Send a command to speak from 15 to DSP13
(Step S266).

In response to this, the DSP 13 utters the synthesized voice from the speaker 15 and transmits an end response signal indicating the end thereof. When the microcomputer 21 receives the end response signal transmitted from the DSP 13 (step S26
7), decrement the variable j by "1" (step S268)
Further, it is determined whether or not the result is smaller than "0" (step S269). The process is returned from step S269 to step S266 until the variable j becomes "0", and therefore the microcomputer 21 repeats the loop process of steps S266 to S269.

By the way, from step S61 to step S62
The variable j represents the number of recognized numbers when the process is advanced to. In step S64, "k =
At this point, the variable k also represents the number of recognized numbers, and in the first loop processing of step S266, "k =
Since it is j ”, it is“ 0 ”th, in other words, the variable j is“
The command for uttering the number recognized at the time of "0" with the synthesized voice is transmitted from the microcomputer 21 to the DSP 13. After that, the above steps S266 to S2 are performed.
Each time the loop processing of 69 is repeated, the variable j is decremented by "1" at step S268, and therefore, at step S266, the number recognized at each time when the variable j is "1", "2" ... Will be uttered from the speaker 15 as a synthetic voice.

Eventually, when the variable j becomes smaller than "0", the process exits from the loop process, the microcomputer 21 clears both variables j and i to zero (step S270), and returns the process to step S32. Through the above processing, each number in the recognized number string is sequentially and continuously uttered as a synthetic voice from the speaker 15.

The other processing is the same as that of the first embodiment described above, and the description thereof is omitted.

As described above, in the third embodiment of the G code remote controller as the numeral string voice recognition apparatus of the present invention, the confirmation of the recognition result of the numeral inputted by voice is confirmed as in the first and second embodiments. Instead of asking the user each time, all the numbers are input and then collectively uttered as a synthetic voice. Due to these characteristics,
The third embodiment of the present invention enables quick input as compared with the first and second embodiments described above. This will be specifically described below.

Now, for example, as in the case of the above-mentioned conventional example, when the user inputs the numeral string "5492" by voice, the response of the third embodiment of the G code remote controller as the voice recognition apparatus of the present invention is responsive to it. This will be described with reference to the schematic diagram of FIG.

In the third flow chart shown in FIG. 12, FIG. 13 and FIG. 14 described above, when a certain voice captured by the DSP 13 is recognized, only confirmation that the recognition is performed is given in step S142. A short confirmation sound such as "beep" is emitted from the speaker 15. That is, the figure
As shown in FIG. 15, when the user first says “go”, the speaker 15 beeps, and when the user says “yeon”, the speaker 15 beeps, Next, when the user utters "kyu", the speaker 15 beeps, and when the user finally utters "d", the speaker 15 beeps. When the user turns off the G code button 23, the speaker 15 displays
The recognition result is continuously uttered as "Yeon Kyu Ni".

In the third embodiment of the G code remote controller as the number string voice recognition device of the present invention, a microcomputer is used.
21 is programmed to perform the above-mentioned processing, but the procedure for the user to actually use the third embodiment of the G code remote controller as the numeral string voice recognition device of the present invention will be described below. To do.

Before the user starts using the G code remote controller as the number string voice recognition device of the present invention, the microcomputer 21 is in a state of waiting for some operation by the user. That is, since both the transfer button 22 and the G code button 23 are off,
The process proceeds in the order of S31, S32, S71, S78, and then steps
Returning to S32, the loop processing is repeated in the order of steps S71, S78 and S32.

When the user keeps the G code button 23 in the ON state to input the G code, the ON signal is given to the microcomputer 21. Microcomputer 21
Detects an ON signal from the G code button 23 in step S32 during the loop processing of steps S32, S71, S78, S32. As a result, the microcomputer 21 thereafter proceeds from step S32 to step S35, and enters a loop process that reciprocates between step S35 and step S61. When the user utters the first digit of the G code in this state, the DSP 13 captures the user's voice. After that, the microcomputer 21 executes the processing of each step from step S35 to step S45.
Recognizes the previously captured voice.

In this way, the microcomputer 21 repeats the loop processing of sequentially returning from step S35 to step S45 and returning to step S32, whereby the G code of an arbitrary number of digits sequentially uttered by the user. For each loop processing, each number is captured one by one and the reject threshold is “reject” = 150, score ratio threshold ”
The voice is recognized with the reference value of score "= 0.5, and a confirmation sound such as" beep "is emitted from the speaker 15 every time one number is recognized.

If the individual number recognition process is rejected during the process, step S39 or step S41 is performed.
From step S46 to S47, a reject sound such as "buzz" is uttered.

When all the numbers of the G code have been input, the user turns off the G code button 23. Then
The recognized G-code number sequence is uttered from the speaker 15 in sequence as a synthetic voice, so the user confirms whether or not the desired number sequence is input, and then the transfer button 22 is turned on. do it. Subsequent operations are the above-mentioned first
This is the same as the embodiment. If the G code is not entered as the user desires, press the G code button again.
When 23 is turned on, the user can newly input the G code again.

In the third embodiment, the reject threshold "reject" is set to "150" and the score ratio threshold "score" is set.
Is set to "0.5" for voice recognition, but it goes without saying that other suitable numerical values may be adopted as in the first embodiment.

[Fourth Embodiment] Next, regarding the fourth embodiment of the G code remote controller as the numeral string voice recognition device of the present invention, the processing procedure of the microcomputer 21, in other words, the voice recognition of the present invention. 16 and 17 showing the method divided
18 and the flowchart of FIG. 18. In addition,
Hereinafter, the flowcharts shown in FIGS. 16, 17, and 18 will be referred to as a fourth flowchart. Also, in FIG.
A first showing the processing procedure of the first embodiment among the processing steps of the fourth flowchart shown in FIGS. 17 and 18.
Flow chart of the second embodiment showing the processing procedure of the second embodiment
Flow chart of the third embodiment showing the processing procedure of the third embodiment
The same step numbers are given to the same portions as the processing steps in the flowchart of FIG.

In the fourth flow chart shown in FIGS. 16, 17 and 18, step S31 in the first flow chart shown in FIG. 4, FIG. 5 and FIG. 6 is replaced with step S31. S131 replaces step S38 with step S1
38, step S144 instead of step S44
Steps S265 to S270 are executed instead of S65, and the first step shown in FIGS. 4, 5 and 6 is executed.
The step S48 is additionally executed after the step S47 in the flowchart of FIG.

The features of the fourth flow chart shown in FIGS. 16, 17 and 18, that is, the features of the fourth embodiment are, in short, the features of the second embodiment and the features of the third embodiment. This is a feature that combines the features of the embodiment. That is, the reject threshold "reject" at the time of re-recognition when the recognition result of the numbers input by voice, which is the feature of the second embodiment, is rejected.
And the point that the score ratio threshold “score” is relaxed, in other words, the possibility of being rejected after re-recognition once rejected is lower than that of the previous time, and the feature of the third embodiment. That is, the recognition result of each number inputted by voice can be collectively confirmed at the time when the input of the number sequence is completed.

The features of the fourth embodiment as described above are mainly realized by the processing of step S138 and the processing of steps S156 to S270. That is, the microcomputer 21 first clears the variable i indicating the number of times the rejects are consecutively performed together with the other variables j and k to zero in step S131, starts the processing in step S32 and thereafter, and loops in the order of steps S32, S71, S78. Start processing. Then, in the recognition processing in step S138 after capturing a certain number of voices, the variable i is "0" at the first time, and therefore, in the same manner as the processing in step S38 of the first embodiment,
Set the reject threshold "reject" = 150 and the score ratio threshold "score" = 0.5, and then step
In S39, S40, S41, it is determined whether or not to reject.

If the reject is not made in this first processing, the variable j is incremented by "1" in step S144 instead of step S44 of the first flowchart, and the variable i representing the number of times the reject is consecutively cleared to zero. It However, if rejected, the variable i is incremented by "1" in step S48 following step S47.

As described above, when rejected in the first recognition process, the variable i is incremented by "1" and becomes "1". Therefore, in the next step S138, the rejection threshold value "reject" = 100 and the score is again set. Ratio threshold "score"
= 0.8 respectively and then steps S39, S40, S4
At 1, it is judged whether or not to reject.
Therefore, in the second recognition process after the first rejection, it is determined whether or not to reject in a state in which the possibility of rejection is smaller than that in the first recognition process.

Further, in the case where the variables are rejected twice in succession, the variable i is further incremented by "1" in step S48 to become "2". Therefore, in step S138, the rejection threshold "reject" = 0 and the score Ratio threshold ”
score ”= 1.0, and then step S39,
Whether or not to reject is determined in S40 and S41. Therefore, at the time of the third recognition process after the second successive rejection, it is determined whether or not to reject in the state where there is no possibility of rejection. In other words, the recognition candidate of the first-ranked recognition score is unconditionally determined as the recognition result after being rejected twice in succession.

When the voice input of the G code is completed or is stopped midway, the recognized digit is obtained by the processing of steps S265 to S270 instead of step S65 shown in the second flowchart. The rows are collectively uttered from the speaker 15 as synthetic voice. That is, the microcomputer 21 first determines whether the value of the variable k, that is, the number of recognized numbers is 1 or more, in other words, whether or not even one number is recognized (step S2
65).

If it is determined in step S265 that no numeral has been recognized, the microcomputer 21 returns the process to step S32 and enters the standby state.
If at least one number is recognized in step S265, the microcomputer 21 sends to the DSP 13 a command for producing a synthetic voice corresponding to the recognition result of the "kj" th recognized voice. (Step S266)
.

In response to this, the DSP 13 utters the synthesized voice from the speaker 15 and transmits an end response signal indicating the end thereof. When the microcomputer 21 receives the end response signal transmitted from the DSP 13 (step S26
7), decrement the variable j by "1" (step S268)
Further, it is determined whether or not the result is smaller than "0" (step S269). The process is returned from step S269 to step S266 until the variable j becomes "0", and therefore the microcomputer 21 repeats the loop process of steps S266 to S269.

By the way, from step S61 to step S62
The variable j represents the number of recognized numbers when the process is advanced to. In step S64, "k =
Since it is set as j ″, the variable k also represents the number of recognized numbers at this time.
The first step S2 of the loop processing of 66 to step S269
In the process of 66, since it is “k = j”, it is the “0”.
In other words, in other words, a command for uttering the number recognized when the variable j is "0" with synthetic voice is transmitted from the microcomputer 21 to the DSP 13. After that, each time the above loop processing is repeated, the variable j is decremented by "1" in step S268, and thus step S2 is performed.
In FIG. 66, the numbers recognized at the time when the variable j is “1”, “2” ...
Will be uttered by.

Eventually, when the variable j becomes smaller than "0", the process exits from the loop process, the microcomputer 21 clears both variables j and i to zero (step S270), and returns the process to step S32. Through the above processing, each number in the recognized number string is sequentially uttered as a synthetic voice from the speaker 15.

The other processing is the same as that of the first embodiment described above, and therefore its explanation is omitted.

As described above, in the fourth embodiment of the G code remote controller as the numeral string voice recognition device of the present invention, the second code is used.
It has both the features of the embodiment of (3) and the features of the third embodiment. Therefore, it is possible to prevent the frequent occurrence of erroneous recognition, the frequent occurrence of rejects, and smooth voice input.

In the fourth embodiment of the G code remote controller as the numeral string voice recognition device of the present invention, a microcomputer is used.
21 is programmed to perform the above-mentioned processing, but the procedure for the case where the user actually uses the fourth embodiment of the G code remote controller as the numeral string voice recognition device of the present invention will be described below. To do.

Before the user starts using the G code remote controller as the number string voice recognition device of the present invention, the microcomputer 21 is in a state of waiting for some operation by the user. That is, since both the transfer button 22 and the G code button 23 are off,
The process proceeds in the order of S131, S32, S71, S78, and thereafter, the process returns to step S32 and the loop process is repeated in the order of steps S71, S78, S32.

When the user keeps the G code button 23 in the ON state to input the G code, the ON signal is given to the microcomputer 21. Microcomputer 21
Detects an ON signal from the G code button 23 in step S32 during the loop processing of steps S32, S71, S78, S32. As a result, the microcomputer 21 thereafter proceeds from step S32 to step S35, and enters a loop process that reciprocates between step S35 and step S61. When the user utters the first digit of the G code in this state, the DSP 13 captures the user's voice. After that, the microcomputer 21 executes the processing of each step from step S35 to step S45.
Recognizes the voice previously captured and utters the speaker 15 with the synthesized voice of the recognition result.

As described above, the microcomputer 21 repeats the loop processing of sequentially returning from step S35 to step S45 and returning to step S32, whereby the G code of an arbitrary number of digits uttered by the user sequentially. For each loop processing, each number is captured one by one and the reject threshold is “reject” = 150, score ratio threshold ”
Speech is recognized with a reference value of "score" = 0.5, and the synthesized speech of the recognition result is uttered from the speaker 15.

If the individual number recognition process is rejected midway, step S39 or step S41 is performed.
From step S46 to S47, a reject sound such as "buzz" is uttered, the variable i is incremented by "1" in step S48, and then step S3.
The process is returned to 2. Therefore, in this case, if the user continues to turn on the G code button 23, the variable i becomes "1", and thus the reject threshold "reject" = 100 and the score ratio threshold "score" = 0.8, which is more than the first time. Re-recognition is possible when the reject reference value is low. But,
When rejected twice in succession, the variable i is "
When it becomes 2 ”, the rejection threshold“ reject ”= 0 and the score ratio threshold“ score ”= 1.0, and the recognition candidate of the first recognition score is unconditionally determined as the recognition result. Therefore, the rejection is repeated. The possibility that the input voice is reduced is recognized, and at the maximum, the third input voice after two consecutive rejects is recognized.

After inputting all the numbers of the G code, the user turns off the G code button 23. Then
The recognized G-code number sequence is uttered by the synthesized voice sequentially and continuously from the speaker 15. Therefore, the user confirms whether or not the desired number sequence is input, and then the transfer button. Turn on 22. The subsequent operations are the same as those in the first embodiment described above. If the G code has not been input as the user desires, the G code button 23 is turned on again to allow the user to newly input the G code again.

In the fourth embodiment, the reject threshold "reject" is set to "150" every time reject is performed.
"," 100 "," 0 ", score ratio threshold" score "
Are sequentially changed to "0.5", "0.8", "1" in three steps, but even if they are changed to two steps as in the second embodiment described above, or to four steps or more. You can change it,
Needless to say, numerical values other than the above may be adopted.

In each of the above embodiments, the voice recognition method and the number string voice recognition device of the present invention are applied to the G code remote controller for G code voice input. However, the present invention is not limited to this, and other examples such as It is needless to say that the present invention can be applied to various devices for inputting a number string composed of a plurality of digits, such as a device for inputting a voice of a telephone number.

[0145]

As described above in detail, according to the voice recognition method and the numerical string voice recognition device of the present invention, the recognition candidate having the first-ranked recognition score which is generally used in the past is directly recognized. The resulting method, or if the first-ranked recognition score is equal to or higher than the reference value, the recognition candidate is not used as the recognition result, and the ratio of the second-ranked recognition score to the first-ranked recognition score is further added. When is less than or equal to the reference value, in other words, when there is a sufficiently significant difference between them, the recognition candidate having the first recognition score is set as the recognition result. As a result, it is rejected in an ambiguous case in which there is no significant difference between the recognition scores of two or more recognition candidates, and the possibility of erroneous recognition decreases.

Further, according to the voice recognition method and the numerical string voice recognition apparatus of the present invention, when re-recognizing once a reject occurs, the reject criterion is relaxed, so that rejects do not occur frequently. The user is freed from the trouble of having to repeat the same utterance many times.

Further, according to the number string voice recognition apparatus of the present invention, every time the user utters a single number, a response voice indicating only that the recognition is performed is emitted, and the user recognizes all the numbers. When the input is completed, all the numbers recognized so far are sequentially and continuously uttered as synthesized speech. As a result, the user can sequentially recognize the numbers by uttering the numbers at his own pace, so that the safety can be secured especially when applied to, for example, an automobile telephone.

Furthermore, according to the video recorder system of the present invention, by using the above-mentioned number string voice recognition device as a remote control device, it is possible to make a reserved recording of the video recorder main body by voice input. This makes it easy for people who are not good at operating machines.

[Brief description of drawings]

FIG. 1 is a schematic diagram showing the principle of a voice recognition method according to the present invention.

FIG. 2 is a block diagram showing a hardware configuration example of a G code remote controller as a number string voice recognition device according to the present invention.

FIG. 3 is a schematic diagram showing an appearance of a G code remote controller as a number string voice recognition device of the present invention.

FIG. 4 is a flowchart showing an operation procedure of the first embodiment of the G code remote controller as the numeral string voice recognition device of the present invention.

FIG. 5 is a flowchart showing an operation procedure of the first embodiment of the G code remote controller as the numeral string voice recognition device of the present invention.

FIG. 6 is a flowchart showing an operation procedure of the first embodiment of the G code remote controller as the numeral string voice recognition device of the present invention.

FIG. 7 is a view for explaining an operation state of the first embodiment of the G code remote controller as the numeral string voice recognition device of the present invention;
6 is a graph showing a distribution of recognition scores when a speaker utters "d".

FIG. 8 is a flowchart showing an operation procedure of the second embodiment of the G code remote controller as the numeral string voice recognition device of the present invention.

FIG. 9 is a flowchart showing an operation procedure of the second embodiment of the G code remote controller as the numeral string voice recognition device of the present invention.

FIG. 10 is a flowchart showing an operation procedure of the second embodiment of the G code remote controller as the numeral string voice recognition device of the present invention.

FIG. 11 is a graph showing the distribution of recognition scores for explaining the operation state of the second embodiment of the G code remote controller as the number string voice recognition device of the present invention.

FIG. 12 is a flowchart showing an operation procedure of the third embodiment of the G code remote controller as the numeral string voice recognition device of the present invention.

FIG. 13 is a flowchart showing an operation procedure of the third embodiment of the G code remote controller as the numeral string voice recognition device of the present invention.

FIG. 14 is a flowchart showing an operation procedure of the third embodiment of the G code remote controller as the numeral string voice recognition device of the present invention.

FIG. 15 is a schematic diagram showing a response state to a voice input of the third embodiment of the G code remote controller as the voice recognition device of the present invention.

FIG. 16 is a flowchart showing an operation procedure of the fourth embodiment of the G code remote controller as the numeral string voice recognition device of the present invention.

FIG. 17 is a flowchart showing the operation procedure of the fourth embodiment of the G code remote controller as the numeral string voice recognition device of the present invention.

FIG. 18 is a flowchart showing an operation procedure of the fourth embodiment of the G code remote controller as the numeral string voice recognition device of the present invention.

FIG. 19 is a schematic diagram showing the concept of the principle of voice recognition by a conventional voice recognition device.

FIG. 20 is a flowchart showing a specific processing procedure by a conventional voice recognition device.

FIG. 21 is a schematic diagram showing a response state of a conventional voice recognition device when a number string “5492” is input by voice.

FIG. 22 is a schematic diagram showing a response state of a conventional voice recognition device when a number string “5492” is input by voice.

[Explanation of symbols]

 10 DSP section 11 Microphone 13 DSP 15 Speaker 16 PROM 18 RAM 19 ROM 20 Microcomputer section 21 Microcomputer 22 Transfer button 22 23 G code button 25 LCD 40 VTR

Claims (11)

[Claims] 1. A recognition score indicating the accuracy of a recognition result for a speech signal to be recognized is obtained for each of a plurality of recognition candidates, and the first-ranked recognition score among the obtained recognition scores is first.
The first determination is performed by comparing the first recognition score with the first recognition score as a result of the first determination.
If it is smaller than the reference value of, all recognition candidates are rejected, and as a result of the first determination, the first recognition score is the first recognition score.
If it is larger than the reference value of, the score ratio representing the ratio of the recognition score of the second place to the recognition score of the first place is obtained, and the obtained second score is compared with the second reference value. If, as a result of the second determination, the score ratio is larger than the second reference value, all recognition candidates are rejected, and as a result of the second determination, the score ratio is the second. A voice recognition method, characterized in that a recognition candidate for which a first-ranked recognition score is obtained is regarded as a recognition result when the recognition candidate is smaller than the reference value of. 2. When the same object to be recognized is rejected according to the result of the first determination or the second determination, the first reference value is set to a smaller value and the second reference value is set to a smaller value.
The voice recognition method according to claim 1, wherein the reference value is set to a larger value and the first determination and the second determination are performed again. 3. When the rejection according to the result of the first judgment or the second judgment for the same recognition target is continuously performed a predetermined number of times, the first reference value is set to "
The voice recognition method according to claim 2, wherein the second reference value is set to "0" and the second reference value is set to "1". 4. A recognition score calculation means for calculating, for each of a plurality of recognition candidates, a recognition score indicating the accuracy of a recognition result for a voice signal of each one-digit number of a series of numbers to be recognized, the recognition score calculation means. A first determination unit that compares the first-ranked recognition score with a first reference value; and, as a result of the determination by the first determination unit, the first-ranked recognition score is the first reference value. When the first recognition score is larger than the first reference value as a result of the judgment by the first rejecting means for rejecting all the recognition candidates when it is smaller than the above, and the first judging means. Is a score ratio calculation means for obtaining a score ratio representing the ratio of the second-ranked recognition score to the first-ranked recognition score calculated by the recognition score calculation means; and the score calculated by the score ratio calculation means. The second
Second determination means for comparing with the reference value of, and as a result of the determination by the second determination means, if the score ratio is larger than the second reference value, reject all recognition candidates. As a result of the judgment by the rejecting means and the second judging means, if the score ratio is smaller than the second reference value, the recognition candidate for which the first recognition score is calculated by the recognition score calculating means. And a recognition result output means for outputting as a recognition result. 5. When the same rejected object is rejected by the first reject means or the second reject means, the first reference value is set smaller and the second reference value is set smaller. 5. The number string voice recognition device according to claim 4, further comprising re-determination means which is set to a larger value and causes the first determination means and the second determination means to perform the determination again. 6. The re-determining means sets the first reference value when the first rejecting means or the second rejecting means rejects the same recognition target continuously a predetermined number of times. 6. The number string voice recognition device according to claim 5, wherein the second reference value is set to "0" and the second reference value is set to "1". 7. A voice output means, and each time the recognition result output means outputs a recognition result, the voice output means is caused to utter a predetermined synthetic voice for all numbers in a series of numbers to be recognized. The number according to claim 4, further comprising: a unit that causes the voice output unit to continuously utter each number of the recognition result in a synthesized voice when the recognition result output unit outputs the recognition result. Row voice recognition device. 8. A video recorder system comprising a video recorder main body, which is designed to make a recording reservation by being given data of a predetermined format, and a remote control device for transmitting the data of the predetermined format. Recognition for each of a plurality of recognition candidates, a recognition score indicating the accuracy of the recognition result for the voice signal of each one-digit number of the series of numbers to be recognized that is the source of the data to be given to the video recorder main body. Score calculation means, first determination means for comparing the first-ranked recognition score calculated by the recognition score calculation means with a first reference value, and the first-ranked result as a result of the judgment by the first determination means. A first rejecting means for rejecting all recognition candidates when the recognition score is less than the first reference value; As a result of the determination by the first determination means, if the first-ranked recognition score is larger than the first reference value, the second-ranked one with respect to the first-ranked recognition score calculated by the recognition-score calculating means. The score ratio calculating means for obtaining a score ratio representing the ratio of the recognition scores of the
Second determination means for comparing with the reference value of, and as a result of the determination by the second determination means, if the score ratio is larger than the second reference value, reject all recognition candidates. As a result of the judgment by the rejecting means and the second judging means, if the score ratio is smaller than the second reference value, the recognition candidate for which the first recognition score is calculated by the recognition score calculating means. A recognition result output means for outputting as a recognition result, each number of recognition candidates sequentially determined as a recognition result of one number by the recognition result output means is a number string as data of a predetermined format according to a predetermined logic The remote control device is provided with a converting means for converting into the video recording device, and a transmitting means for transmitting the data obtained by the converting means to the video recorder main body. Video recorder system 9. When the same rejected object is rejected by the first rejecting means or the second rejecting means, the first reference value is set smaller and the second reference value is set smaller. 9. The video recorder system according to claim 8, wherein the remote control device is provided with a re-judging unit which is set to a larger value and causes the first judging unit and the second judging unit to judge again. . 10. The re-determining means sets the first reference value when the first rejecting means or the second rejecting means rejects the same recognition target continuously a predetermined number of times. 10. The video recorder system according to claim 9, wherein the second reference value is set to "0", and the second reference value is set to "1". 11. A voice output means, and each time the recognition result output means outputs a recognition result, the voice output means utters a predetermined synthetic voice, and with respect to all numbers of a series of number strings to be recognized. 7. The remote control device comprises: means for causing the voice output means to continuously utter each number of the recognition result when the recognition result output means outputs the recognition result. 8. The video recorder system according to item 8.