JP2975633B2 - Voice recognition method - Google Patents

Description

The present invention relates to a speech recognition method in a system for mechanically recognizing speech uttered by a speaker.

(B) Conventional technology As a system for mechanically recognizing a voice uttered by a speaker, various voice recognition devices have been conventionally developed.
In recent years, for example, it has been adopted for telephone systems that can automatically dial by voice (Japanese Patent Laid-Open No. 62-81).
No. 152).

However, although research and development of speech recognition techniques have progressed and false recognition has been reduced, there are uncertain factors beyond human utterance and the effect of ambient noise cannot be avoided. At present, it cannot be completely eliminated.

Therefore, as described above, in the auto-dial telephone system, as a first step, the name of the caller is recognized by voice, and this recognition result is displayed or reported by synthetic voice, and when the reported recognition result is correct. First, in the second step, execution command words such as "dial" are subjected to voice recognition so as to dial the destination name of the recognition result.
When the notified recognition result is an error, a method (two-stage recognition processing method) of performing voice recognition of a non-execution command word such as "cancel" in the second step is adopted.

However, in such a conventional two-stage recognition processing method, the user, that is, the speaker, has to repeat the input of the non-execution command word and the re-input of the destination name every time the destination name is erroneously recognized. Therefore, there is a disadvantage that the burden on the user is increased.

(C) Problems to be Solved by the Invention The present invention has been made in order to solve the above-mentioned drawbacks, and is based on the premise that erroneous recognition may occur, and reduces the user's burden. Is provided.

(D) Means for Solving the Problems The voice recognition method of the present invention includes the following processing steps.

A first voice recognition processing step of recognizing a voice uttered by a speaker to obtain a plurality of recognition candidates; a recognition candidate notification process of sequentially outputting a plurality of recognition candidates obtained by the first voice recognition processing to a speaker A command voice spoken by the speaker to indicate that the candidate corresponds to the speaker of the speaker in the first speech recognition process when a specific recognition candidate is notified in the recognition candidate notification process. A second speech recognition processing step of recognizing the target speech, if the instruction speech has not been recognized by the second speech recognition processing by the time when the notification of all candidates has been completed in the recognition candidate notification processing, A guidance notification processing step of performing guidance notification for asking the speaker whether to execute the candidate notification processing again; in the second voice recognition processing in which the speaker utters in response to the guidance notification in the guidance notification processing; Command word voice that is the same as A third speech recognition processing step of recognizing a recognition candidate re-notification processing step of re-executing the recognition candidate notification processing when the command word voice can be recognized in the third speech recognition processing;

(E) Function In the voice recognition method of the present invention, since a plurality of recognition candidates are sequentially notified and output, even if the recognition result of the first candidate is erroneous, a desired recognition result is selected from among the candidates notified one after another. Then, a command word voice such as “OK” is input to the candidate by voice, and if the command word voice is recognized, the candidate can be output as a correct recognition result. Moreover, if the command word voice cannot be recognized, or if the speaker cannot input the command word voice, the speaker responds to the guidance notification asking for re-notification of these recognition candidates. By performing voice input for recognizing the same command voice as the command voice, if the command word voice can be correctly recognized at this time, the recognizability of the command word voice itself can be confirmed, and at the same time, these recognition candidates are recognized. Re-notification can be performed.

(F) Embodiment An embodiment in which the present invention is applied to a wireless auto-dial telephone mounted on an automobile will be described below with reference to the drawings.

1 and 2 show the electrical configuration of a wireless auto-dial telephone, where FIG. 1 shows a new unit and FIG. 2 shows a slave unit.

2 includes a microphone 1 for inputting voice, an amplifier 2 for amplifying a voice signal input from the microphone 1, a modulation circuit for modulating a carrier with the amplified voice signal, and an output from a modulation circuit. A transmitting unit 3 including an amplifier circuit for amplifying a modulated wave to be received, an antenna coupler 4, a transmitting / receiving antenna 5, an amplifier circuit for modulating a received modulated wave, and a demodulation circuit for demodulating the amplified modulated wave. , An amplifying unit 7 for amplifying the audio signal reproduced by the receiving unit 6, a speaker 8 for outputting the amplified audio signal as audio, a numeric keypad, a key input unit 9 having various function keys, and a transmitting unit 3, a control unit 10 for controlling the receiving unit 6 and the key input unit 9 is provided.

The master unit shown in FIG. 1 includes a loop detecting unit 27 for detecting a line loop state, a voice recognizing unit 25 for performing auto dialing by voice input, and a telephone number corresponding to the voice recognized by the voice recognizing unit 25. Line control unit 2 that supplies transmission / reception signals to the line while supplying transmission / reception signals to the line.
1, an antenna coupler 22, a receiving unit 23 including an amplification circuit for amplifying a received modulated wave and a demodulation circuit for demodulating the amplified modulated wave, a line control unit 26 for transmitting an audio signal reproduced by the receiving unit 23. And a changeover switch 24 for switching which of the voice recognition units 25 to supply, a numeric keypad,
A key input unit 34 having various function keys, a voice synthesis unit 28 for forming various synthesized voices, a modulation circuit and a modulation circuit for modulating a carrier by an output signal of the voice synthesis unit 28 or a reception signal from the line control unit 26 A transmission unit 29 including an amplifier circuit for amplifying a modulated wave output from the circuit and a control unit 30 for controlling these devices are provided.

The voice recognition unit 25 includes an analysis unit 41 that performs frequency analysis of an input voice signal, a time series group of a voice spectrum based on an analysis result of the analysis unit 41, that is, a voice pattern generation unit 42 that generates a voice pattern, and a pattern generation unit. An input voice pattern memory 45 for storing the voice pattern created by 42, a first reference pattern memory 43 for storing standard voice patterns of a plurality of call destination names, and a standard voice pattern for a plurality of command words. An identification unit 46 is provided for comparing the audio pattern output from the second standard pattern memory 44 and the audio pattern creation unit 42 with the standard pattern and identifying the similarity between the two patterns.

The control unit 30 has a main memory 31 for storing programs and necessary data,
A recognition result memory 32 for storing caller names in descending order of similarity and a counter 33 for time measurement are provided.

Further, the control unit 30 includes a detection signal of a steering wheel rotation angle detector 35 for detecting a rotation angle of a steering wheel of an automobile on which the telephone is mounted, and a detection signal of a brake operation detector 36 for detecting an operation state of a brake. And a detection signal of a vehicle speed detector 37 for detecting the vehicle speed is input.

In order to perform auto-dial by voice using this wireless auto-dial telephone, voice recording of the command word and the name of the called party is performed in advance. Examples of the command word include a command word “off hook” for making the telephone go off-hook and a command word “OK” for confirmation.

The voice registration of the command word and the call destination name is performed as follows. First, a registration key provided on the key input unit 9 of the slave unit is pressed. Then, the control unit 10 of the slave unit outputs a registration operation start signal. Then, in the transmission unit 3,
The carrier is modulated by the registration operation start signal, and the modulated wave is transmitted via the coupler 4 and the antenna 5.
This modulated wave is received by the parent device Anten 21,
The signal is sent to the receiving unit 23 via the coupler 22. And the receiving unit
At 23, the received modulated signal is demodulated and amplified. The registration operation start signal reproduced by the receiving unit 23 is transmitted to the control unit.
30 to enter a registration operation state.

Instead of pressing the registration key of the key input unit 9 of the child device, the registration key of the key input unit 34 of the parent device may be pressed. In this case,
When the output of the key input unit 34 is input to the control unit 30 of the master unit, a registration operation state is set.

Next, the command word and the call destination are input by the microphone 1 of the slave unit. The audio signal output from the microphone 1 is sent to the transmission unit 3 after being amplified by the amplification unit 2. Then, in the transmission unit 3, the carrier is modulated by the audio signal, and the modulated wave is amplified. The modulated wave output from the transmitting unit 3 is transmitted via the coupler 4 and the antenna 5.

This modulated wave is received by the antenna 21 of the master unit, and sent to the receiving unit 23 via the coupler 22. Then, in the receiving unit 23, the received modulated wave signal is demodulated,
Amplified. The audio signal reproduced by the receiving unit 23 is sent to the audio recognizing unit 25 via the changeover switch 24.

In the voice recognition unit 25, first, the input voice signal is subjected to frequency analysis by the voice analysis unit 41, and a voice pattern is generated by the pattern generation unit 42 based on the analysis result. This voice pattern is stored in the first standard pattern memory when the input voice is the call destination name, and is stored in the second standard pattern memory when the input voice is the command word.

By such an operation, the voice patterns of all call destination names and the voice patterns of all command words are stored in the first or second standard pattern memory.

At the time of voice registration of the call destination name, following the voice input of each call destination name, the telephone number of the call destination is input using the numeric keypad of the key input sections 9 and 34 of the master unit or the slave unit, and the voice registration is performed. The telephone numbers corresponding to the respective called parties are stored in the main memory 31.

FIGS. 3 (I) to (IV) show a procedure of a calling process by the control unit 30 of the master unit.

First, after turning on the power of the slave unit, the user (speaker) inputs the off-hook command word “off-hook” by using the microphone 1. When the off-hook command word is input to the microphone 1 (step S1), the audio signal from the microphone 1 is amplified, the carrier wave is modulated by the amplified audio signal, and the modulated wave is transmitted from the antenna 5 of the slave unit. Sent. This modulated wave is received and demodulated by the antenna 21 of the master unit. The reproduced voice signal is sent to the voice recognition unit 25 via the changeover switch, and the voice recognition processing is performed (step S2).

In the voice recognition unit 25, first, the voice analysis unit 41 performs frequency analysis on the input voice signal. Next, a voice pattern is generated by the pattern generation unit 42 based on the analysis result, and the generated voice pattern is stored in the input voice pattern memory 45. Then, in the identification unit 46, the created voice pattern and the second standard pattern memory are stored.
All command word standard patterns stored in 44 are compared, and their similarity is identified.

When the input voice is recognized as "off-hook" by the identification unit 46 (step S3), a DC loop is formed by the line control unit 26, and an off-hook state is set (step S4).
Thereafter, a safety confirmation process is performed to check whether the driving situation is in a safe state with respect to the user performing the telephone operation (step S5). Details of this processing will be described later.

In the safety confirmation process, if it is confirmed that the security is secured, the voice synthesizer 28 prompts the user to input the name of the called party, for example, "Please enter the called party name." Is output. This signal is sent to the transmission unit 29, and the carrier is modulated by this signal. The modulated wave is transmitted via the antenna 21 after being amplified.

This modulated wave is received by the antenna 5 of the slave unit,
Demodulated. The reproduced speech synthesis signal is sent to the speaker 8 after being amplified, and the speaker 8 outputs a voice saying "Please input the name of the other party" (step S).
6).

Thereafter, the user enters the name of the called party, for example, "Sanyou"
Is input from the microphone 1 (step S7).
The voice signal is received by the base unit by wireless communication, and the voice recognition unit 25 calculates the similarity to all the partner standard patterns in the first standard pattern memory 43. Then, in the recognition result memory 32 in order from the one having the highest similarity,
The call destination name is stored (step S8). Thereafter, the candidate rank n is updated by 1 (step S9). Note that the candidate rank n is set to n = 0 in the initial setting, so that n = 1 at the first time.

Thereafter, a safety confirmation process is performed (step S10).
In the safety confirmation process, when it is confirmed that the call is secure, the n-th candidate among the call destinations stored in the confirmation result memory 32 is read. At the first time, the speech synthesis unit 28 outputs the first candidate, that is, the synthesized speech signal corresponding to the call destination name having the highest similarity. Based on this output, the n-th candidate call destination name is output from the speaker 8 of the slave unit (step S11).

Further, a measurement operation for measuring a predetermined time is started (step S12). This predetermined time is set to an appropriate time (for example, 3 seconds) during which the user can input the confirmation command word “OK” for the voice output of the call destination name. The user can confirm the command word “O” only when the name of the called party whose voice is output is the name of the calling party that is spoken by the user.
"K" is input from the microphone 1 of the slave unit.

When the user voice-inputs “OK” from the microphone 1 of the child device within the above-mentioned predetermined time (step S13), the voice signal is received by the parent device by wireless communication, and the voice recognition unit 25 changes the input voice pattern. It is compared with all the command word standard patterns in the second standard pattern memory 44 (step S1).
Five). As a result of the voice recognition, when the input voice is recognized as "OK" (step S16), the telephone number corresponding to the name of the call destination output in step S11 is stored in the main memory 31.
And sent to the line via the line controller 26 (step S17). As a result, a call is made to the called party.

When the loop detector 27 detects that the called party has gone off-hook within a certain time (step
S18), the changeover switch 24 is switched so that the audio signal input to the slave unit is supplied to the line control unit 26 (step S19). As a result, a call can be made with the call destination. When the call ends (step S20), the contents of the recognition result memory 31 are cleared and the candidate rank n
Is reset (n = 0) (step S21), and this process ends.

If there is no voice input within a predetermined time after the start of the timing operation in step S21 (step S14), and if the input voice is not recognized as "OK" even if there is voice input (step S14). S16) includes the above step S
It is checked whether or not the call destination name output as a voice in step 11 is the final candidate (n = N) (step S22).

If it is not the final candidate (n <N), step S9
And the candidate rank n is updated by one. Thereafter, a safety confirmation process is performed (step S23). When the safety is confirmed, the n-th candidate, that is, the name of the call destination of the next candidate is read from the recognition result memory 32. The synthesized voice signal corresponding to the call destination is output from the voice synthesis unit 28, and based on this, the synthesized voice is output from the speaker 8 of the slave unit (step S11). Further, a timing operation is started (step S12).

Then, for a predetermined period of time, the input of the confirmation command word “OK” is waited for (steps S13 and S14). In the voice input waiting state, there is a voice input, and as a result of the voice recognition,
When the input voice is recognized as the confirmation command word "OK" (steps S13, S15 and S16), processing for auto dialing and talking (steps S17 to S21) is performed.

When there is no voice input in the voice input waiting state of the confirmation command word "OK" (step S14), and when there is a voice input, the input voice is not recognized as the confirmation command word "OK" In (Step S16), it is checked whether or not the destination name that has been voice-output this time is the final candidate (Step S2).
2). Then, if the destination name that has been voice-output this time is not the final candidate (n <N), the process returns to step S9.

If the destination name output this time is the final candidate (n
= N), after the safety confirmation processing is performed (step S23),
A guidance voice for asking the user whether or not to output the destination name again in the candidate order, for example, "Do you want to output the destination name again?" Is output as voice (step S24). In addition, clocking of a predetermined time is started (Step S25). The predetermined time is the same as the confirmation command in the steps S12 and S13, with respect to the guidance voice output,
In this case, the time is set to an appropriate time for inputting “OK”. The user inputs the confirmation command word "OK" by voice from the microphone 1 of the slave unit only when the user desires to output the call destination name again. When the user voice-inputs "OK" from the microphone 1 of the child device within this predetermined time (step S26), the voice signal is received by the parent device by wireless communication, and the second standard It is compared with all the command word standard patterns in the pattern memory 44 (step S28).

As a result of the voice recognition in step S28 in the voice recognition unit 25, if the input voice is recognized as “OK” (step S28).
29), the candidate rank n is reset (step S30).
Then, returning to step S9, processing such as outputting the name of the call destination of the next candidate, waiting for the confirmation command word voice input, and outputting the telephone number when the confirmation command word voice is input is performed again.

If there is no voice input within a predetermined time after the start of the timing operation in step S25, and if the input voice is not recognized as "OK" even if there is a voice input (YES in step S27) Or NO in step S29), the contents of the recognition result memory 32 are cleared and the candidate rank n
Is reset (step S31), and this process ends.

The processing of steps S24 to S30 of the above embodiment is a feature of the present invention. Specifically, in response to voice guidance "Do you want to output the destination name again?" The point is that the same "OK" as the confirmation command word for the voice output of the name is input, and this is recognized.

In other words, even when "OK" is uttered for a desired call destination name during voice output of the call destination names, which are a plurality of recognition candidates in the first round, this may not be recognized due to ambient noise. Therefore, if "OK" for "Do you want to output the destination name again?" Is recognized, "OK" will be displayed even during the voice output of a plurality of destination names in the second round. It can be regarded as indicating that recognition is likely. Such recognition processing of the command word “OK” serves both as a test of the recognizability of the word itself and a command to output the voice of the call destination again.

FIG. 4 shows details of the safety confirmation process. In this process, first, it is determined whether or not the detected rotation angle detected by the steering wheel rotation angle detector 35 is smaller than a predetermined reference angle (step S41).

If the detected rotation angle is smaller than the reference angle, it is next determined whether or not the brake is off based on the output of the brake operation detector 36 (step S42).

If the brake is off, it is determined whether or not the speed detected by the speed detector 37 is lower than a predetermined reference speed (step S43).
If the detected speed is lower than the reference speed, the process ends, and a step S for audio output following the process is completed.
6, the process proceeds to S11 or S24 (see FIG. 3).

If the detected rotation angle is larger than the reference angle in step S41, if the brake is on in step S42, or if the detected speed is larger than the reference speed in step S43, the process returns to step S41.

Therefore, until the detected rotation angle is smaller than the reference angle and the brake is off, and the detected speed is smaller than the reference speed, in steps S6, S11 or S24 (see FIG. 3) following the safety confirmation process. Voice output is prohibited.

In step 2 of the above embodiment, the input voice pattern is compared with all the command word standard patterns stored in the second standard pattern memory 44, but the input voice pattern is stored in the second standard pattern memory 44. It may be compared with only the standard pattern of the off-hook command word.

Further, the destination standard voice pattern and the command word standard voice pattern are stored in separate memories 43 and 44, but may be stored in one standard pattern memory.

In step 8, the similarities with all the standard patterns in the first standard pattern memory 43 are calculated. The similarity with all the partner standard patterns within the predetermined range may be calculated.

For example, the destinations are divided into predetermined groups, the standard patterns of the destinations are stored in the first standard pattern memory 43 for each group, and the group name designation data is input by key input, voice input, etc. before starting the telephone operation. To do it. Then, in step 8, the similarity with all the standard patterns of the destination is calculated within the range of the specified group among the standard voice patterns of the destination stored in the first standard pattern memory 43. To

Further, when there are a plurality of users, the standard pattern is stored in the first standard pattern memory 43 for each user. To enter. Then, in step 8, the first standard pattern memory 43
Of the destination standard voice pattern stored in the destination standard pattern, all similarities of the destination standard pattern for the specified user are calculated.

Further, when there are a plurality of users, the standard pattern is stored in the first standard pattern memory 43 for each user and each predetermined group, and the user name specification data and the group name specification are set before starting the telephone operation. Data is input by key input, voice input, or the like. And
In step 8, among the standard voice patterns of the destination stored in the first standard pattern memory 43, the similarity to all of the standard patterns of the specified user and the specified group is calculated. To do.

In the above embodiment, the names of the call destinations from the first candidate to the final candidate are sequentially output as voice, and in the case where there is no voice input of the command word or there is a voice input of the command word,
If it is not recognized as "OK", it asks the user again whether to output the voice of the call destination name from the first candidate (see steps S22 to S24). However, regardless of the user's intention, the output of the call destination name may be repeated until the number of repetitions m of the output of the call destination name from the first candidate to the final candidate reaches the predetermined number M. Good.

The processing procedure in this case is shown by a broken line in FIG.
That is, if it is determined in step S22 that the call destination name that has been sounded this time is the final candidate (n = Nm), the number of repetitions m is updated by 1 (step S51). The number of repetitions m is set to 0 in the initial setting.

Then, it is determined whether or not the number of repetitions m is greater than a predetermined number M (step S52). If the number of repetitions m is smaller than the predetermined number M, the candidate rank n is reset (step S53), and the process returns to step S9.
Therefore, the call destination name is sequentially output from the first candidate, and if the confirmation command word "OK" is input thereto, the corresponding telephone number is output.

Then, the names of the call destinations from the first candidate to the final candidate are output. However, when there is no input of the confirmation command word “OK” and when there is a voice name input, “OK” is confirmed. If not, as shown in FIG. 3 (IV),
The number of repetitions m is updated by 1 (step S51), and it is determined whether the number of repetitions m is greater than a predetermined number M (step S52). If the number of repetitions m is larger than the predetermined number M, the contents of the recognition result memory 32 are cleared, the candidate rank n and the number of repetitions m are reset (step S54), and this process ends.

In the above two embodiments, the command words “off-hook” and “OK” are input by voice to perform off-hook and confirmation, but other command words may be used. In order to perform off-hook and confirmation, a function key for each command may be provided in the key input unit 34 or 9 of the master unit or the slave unit, and off-hook and confirmation may be performed by key input.

FIG. 5 shows a control unit according to still another embodiment of the present invention.
It shows a part of the outgoing call processing procedure of No. 30 and is changed to FIG. 3 (II). Therefore, the processing procedure other than that shown in FIG.
It conforms to the figures (I), (III) and (IV).

In FIG. 5, the same steps as in FIG. 3 (II) include:
The same step numbers are given and the description thereof is omitted. Comparing the processing of this embodiment with the processing of FIG. 3, the only difference is that steps S61 and S62 are added in this embodiment.

In this embodiment, if the confirmation command word "OK" is not voice-inputted to the calling party name whose voice is output, the next candidate is output as in the case of FIG.

However, if the desired caller name is entered again within a certain period of time after the incorrect caller name is input, the voice of the caller name is confirmed based on the voice input. It is automatically performed again (see steps S61 and S62).

That is, in step 11, for example, the name of the call destination of the first candidate is output as voice, but if the name is different from the one desired by the user, when the user again inputs the name of the call destination by voice, the input is performed. The voice pattern of the voice is stored in the input voice pattern memory 45. Then, the input voice pattern is compared with all the command word standard patterns in the second standard pattern memory 44, and their similarities are checked (step S15).

Next, among the similarities between the respective command word standard patterns and the input voice pattern, the highest similarity α is compared with a predetermined destination / command word discrimination reference value αo (step S61). Since there is a voice distance between the call destination name and the command word, when the call destination name is input as described above, the input voice pattern and the command word standard pattern are compared. The similarity α becomes lower than the reference value αo.

Therefore, the process proceeds from step S61 to step S62, where the contents of the recognition result memory 32 are cleared and the candidate rank n is reset. Then, returning to step S8, the voice pattern of the call destination stored in the input voice pattern memory 45 in step S15 is compared with all the standard patterns in the first standard pattern memory 43, and the similarity is determined. The names of the call destinations are stored in the recognition result memory 32 in descending order of
Is stored. Then, the processes in and after step S9 are performed.

In step S11, when the call destination name output by voice is the one desired by the user, and the user inputs the confirmation command word "OK", the similarity with the command word standard pattern is obtained. Since the degree α becomes higher than the reference value αo, step S61
To step S16. Then, in step S16, the input voice is recognized as "OK", and the telephone number corresponding to the name of the called party whose voice has been output this time is output (step S17).

FIGS. 6 (I) and (II) show a part of a call processing procedure of the control unit 30 in still another embodiment of the present invention.
These are changed to FIGS. 3 (I) and 3 (II), respectively. Therefore, processing procedures other than those shown in FIG. 3 are based on FIGS. 3 (III) and (IV).

In FIG. 6, the same steps as those in FIG. 3 are denoted by the same step numbers, and description thereof will be omitted.

In this embodiment, unlike the case of FIG. 3, when voice recognition of the call destination name is performed, only the first candidate call destination name is output. Therefore, the process of updating the candidate rank n in step S9 in FIG. 3 and the process of repeatedly outputting the destination name from step S22 are not executed. In this embodiment, steps S71 to S76 are added.

In the off-hook state, when voice guidance of "Please input the name of the other party" is given in step S6, a time counting operation for a predetermined time is started (step S71). And
If there is no voice input within this predetermined time (step S7
2), off-hook is released (step S73). Then, the process returns to step 1 to be in a state of waiting for the voice input of the off-hook command word.

If there is a voice input within the above time (step S
7) The input voice pattern is compared with all the partner standard patterns in the first standard pattern memory 43 (step
S8).

Next, the highest similarity β among the similarities between each of the other party standard patterns and the input voice pattern is compared with a predetermined noise discrimination reference value βo (step S7).
Four).

When the similarity β is smaller than the reference value βo, it is determined that the voice input is due to noise, and after the off-hook is released (step S73), the process returns to step S1 to be in a state of waiting for the voice input of the off-tsk command word.

In step S74, when the similarity β is larger than the reference value βo, after the security confirmation process (step S10), the name of the call destination having the highest similarity (first candidate) is output as voice (step S11). . In addition, a timing operation for a predetermined time is started (step S12).

If there is no voice input within the predetermined time (step S14), after off-hook is released (step S75), the process returns to step S1 to be in a state of waiting for voice input of an off-hook command word.

If there is a voice input within the predetermined time (step S13), the input voice pattern and the second standard pattern memory 44
Are compared with all the command word standard patterns (step S15).

Next, the highest similarity β among the similarities between each command word pattern and the input voice pattern is compared with a predetermined noise discrimination reference value βo (step S76).

When the similarity β is smaller than the reference value βo, it is determined that the voice input is caused by noise, and after the off-hook is released (step S75), the process returns to step S1 to wait for the voice input of the off-hook command word.

If the similarity β is larger than the reference value βo in the above step S76, it is determined whether or not the input voice is “OK” (step S16). When the input voice is not recognized as “ON”, after the off-hook is released (step S7
5) Returning to step S1, the apparatus enters a state of waiting for a voice input of an off-hook command word.

If it is determined in step S75 that the input voice is “OK”, a telephone number corresponding to the name of the called party whose voice has been output this time is output (step S17).

FIGS. 7 (I) and (II) show a part of a call processing procedure of the control unit 30 according to still another embodiment of the present invention, which are replaced with FIGS. 3 (I) and (II), respectively. . Therefore, processing procedures other than those shown in FIG. 3 are based on FIGS. 3 (III) and (IV).

In FIG. 7, the same steps as those in FIG. 3 are denoted by the same step numbers, and description thereof will be omitted.

In the processing of FIG. 3, the call destination name is output as voice from the first candidate in order based on the recognition result of the voice input.
In this embodiment, the call destination name is not input by voice, but the call destination name is output by voice in a predetermined order.
Therefore, the process of waiting for the input of the destination name in step S7 of FIG. 3 and the speech recognition processing in step S8 are not performed.

In step S9a corresponding to step S9 in FIG. 3, the predetermined order n is updated instead of the candidate order based on the speech recognition result. Therefore,
In step S11a corresponding to step S11 in FIG. 3, the n-th calling party name in a predetermined order is output as voice. Other points are the same as those in FIG. 3, and thus detailed description is omitted.

In this embodiment as well, the command words "off-hook" and "OK" are input by voice for off-hook and confirmation, but other command words may be used. Further, in order to perform off-hook and check, a function key for each command may be provided in the key input unit 34 or 9 of the master unit or the slave unit, and off-hook and check may be performed by key input.

In all of the above-described embodiments, the off-hook may be automatically set when the voice input device (telephone or microphone) is removed from the normal position, as in a normal telephone, or the microphone or the receiver may be set to the normal position. It may be made to go off-hook by key input or voice input while leaving it.

Also, in all the above embodiments, the present invention is applied to a wireless auto-dial telephone, but it is needless to say that the present invention can be applied to a non-wireless auto-dial telephone.

Further, in the above description, “Please input the destination name” is based on the synthesized voice generated by the voice synthesis unit 28.
Of the initial guidance, the names of the call destinations which are a plurality of recognition candidates, and the guidance voice "Would you like to output the name of the destination again?", Are displayed and displayed on the display unit. Alternatively, the configuration may be as follows.

(G) Effects of the Invention According to the speech recognition method of the present invention, as is apparent from the above description, a command word such as "OK" for a desired recognition candidate is used for reporting a plurality of recognition candidates in the first round. May not be recognized due to the ambient noise even if the utterance is pronounced. If a command word such as “OK”, which is the same as the above-mentioned command word for the guidance voice asking for re-notification of the recognition candidate, can be recognized, the following 2 Even during notification of a plurality of recognition candidates at the round, it can be considered that the command word has a high possibility of being recognized. Therefore, in the speech recognition system that employs the speech recognition method of the present invention, the recognition process of the command word combines the test of the recognizability of the word itself and the command of the notification of the recognition candidate again, The operational burden on the speaker can be greatly reduced.

[Brief description of the drawings]

1 and 2 are block diagrams showing the electrical configuration of a master unit of a wireless auto-dial telephone and its slave unit using the voice recognition method of the present invention, and FIGS. 3 to 7 are processing flowcharts. 1 ... microphone, 3 ... transmitter, 4 ... combiner, 5
…… Antenna, 21 …… Antenna, 22 …… Coupler, 23 ……
Receiving part, 25 ... Voice recognition part, 26 ... Line control part, 27 ...
Loop detection unit 31, 31 recognition result memory, 33 counter, 46 identification unit.

Continuation of the front page (56) References JP-A-62-81152 (JP, A) JP-A-64-81998 (JP, A) JP-A-63-118198 (JP, A) JP-A-61-73998 (JP) JP-A-60-241128 (JP, A) JP-A-1-166100 (JP, A) JP-A-1-154100 (JP, A) Patent 2646080 (JP, B2) JP-B-6-8999 (JP, A) JP, B2) JP 8-33741 (JP, B2) JP 1-443960 (JP, B2) JP 6-34188 (JP, B2) WO 89/4035 (WO, A1) (58) Surveyed field (Int. Cl. ⁶ , DB name) G10L 3/00 561

Claims (1)

(57) [Claims] 1. A first speech recognition processing step of recognizing a speech uttered by a speaker to obtain a plurality of recognition candidates, and sequentially outputting a plurality of recognition candidates obtained by the first speech recognition processing to the speaker. A recognition candidate notification processing step, wherein when a specific recognition candidate is notified in the recognition candidate processing, the speaker utters to indicate that the candidate corresponds to the speaker's utterance in the first speech recognition processing. A second voice recognition processing step of recognizing the command word voice to be performed, when the command word voice is not recognized by the second voice recognition processing by the time when notification of all candidates is completed in the recognition candidate notification processing A guidance notification processing step of performing guidance notification for asking the speaker whether or not to execute the recognition result candidate notification processing again; and the second step in which the speaker utters in response to the guidance notification in the guidance notification processing. Same as command word voice in voice recognition processing of A third speech recognition processing step of recognizing the command word voice, and a re-recognition candidate notification for executing the recognition candidate notification processing again when the command word voice can be recognized in the third voice recognition processing. A speech recognition method including a processing step.