JP5375423B2 - Speech recognition system, speech recognition method, and speech recognition program - Google Patents

Description

  The present invention relates to a voice recognition system, a voice recognition method, and a voice recognition program, and more particularly, to a voice recognition system, a voice recognition method, and a voice recognition program in which a user instructs an utterance timing with a button or the like.

  In this type of speech recognition system, it is well known to prevent erroneous operation due to noise or unintentional speech.

  For example, the speech processing apparatus described in Patent Literature 1 first receives input of a designated section designated by an operator as a section to be subjected to speech processing among input speech, and determines a speech section from the input speech. To detect. Next, the voice processing device determines which of the operator and the person other than the operator is the utterer of the utterance based on the input voice. Furthermore, the voice processing device detects a portion where the designated section and the utterance section overlap, and when the overlapping portion is detected, and when the speaker is determined to be a person other than the operator, An utterance section including an overlapping portion is determined as a section for speech processing.

  As a result, the speech processing apparatus described in Patent Literature 1 can appropriately determine the speech segment to be processed according to the speaker, and can reduce the occurrence of erroneous operations. Further, according to the speech processing device described in Patent Document 1, it is possible to detect an erroneous operation by the operator such as the designation of the start of the designated section being delayed from the actual start of utterance or forgetting to press the end instruction button.

JP 2007-264473 A

  However, in the method described in Patent Document 1, since an utterance section is detected from all input voices, an utterance section detection process must always be performed, and a processing load for detecting the utterance section is increased. large. For this reason, the detection processing load of the utterance section may affect the process of detecting a user's erroneous operation, and the accuracy of erroneous operation detection may be reduced.

  In view of the above, an object of the present invention is to provide a voice recognition system that can detect a user's erroneous operation with reduced processing load for detecting an utterance section.

  In order to achieve the above object, the speech recognition system of the present invention comprises an utterance timing instruction acquisition means for acquiring an instruction of an utterance timing by a user including an instruction to start an utterance, an input voice signal, and the utterance timing When an instruction to start speech is acquired by the instruction acquisition means, the voice signal holding means for outputting the held voice signal and the voice signal input thereafter and the voice signal output by the voice signal holding means An erroneous operation for detecting an erroneous operation of a user based on an utterance interval detecting means for detecting an utterance interval, an utterance interval detected by the utterance interval detecting means, and an instruction of an utterance timing acquired by the utterance timing instruction acquiring means Detecting means.

  In addition, the speech recognition method of the present invention acquires a speech timing instruction by a user including a speech start instruction, retains an input speech signal, and retains the speech start instruction when the speech start instruction is acquired. An audio signal and an audio signal input thereafter are output, an utterance interval is detected from the output audio signal, and an erroneous operation of the user is detected based on the utterance interval and the instruction of the utterance timing.

  Furthermore, the speech recognition program according to the present invention includes an utterance timing instruction acquisition step for acquiring an utterance timing instruction by a user including an utterance start instruction in a computer, an input voice signal, and the utterance timing instruction acquisition step. When an instruction to start utterance is acquired by the voice signal holding step, a voice signal holding step for outputting a held voice signal and a voice signal input thereafter, and a voice interval from the voice signal output by the voice signal holding step are determined. An erroneous operation detecting step for detecting an erroneous operation of the user based on the detected speech interval detecting step, the utterance interval detected by the utterance interval detecting step, and the utterance timing instruction acquired by the utterance timing instruction acquiring step; , Execute.

  ADVANTAGE OF THE INVENTION According to this invention, the processing load of speech area detection can be reduced and a user's misoperation can be detected accurately.

It is a hardware block diagram of the speech recognition system 1 concerning the 1st Embodiment of this invention. It is a block diagram which shows the function structure of the speech recognition system 1 concerning the 1st Embodiment of this invention. It is an illustration of a method for determining the presence / absence / type of an erroneous operation in the erroneous operation detection means. 3 is a flowchart showing the operation of the voice recognition system 1. It is a block diagram which shows the function structure of the speech recognition system 2 concerning the 2nd Embodiment of this invention. It is a block diagram which shows the function structure of the speech recognition system 3 concerning the 3rd Embodiment of this invention. It is a block diagram which shows the function structure of the speech recognition system 4 concerning the 4th Embodiment of this invention.

<First Embodiment>
A first embodiment of a voice recognition system according to the present invention will be described.

  FIG. 1 is a hardware configuration diagram of a speech recognition system 1 according to the first embodiment of the present invention.

  As shown in FIG. 1, a speech recognition system 1 includes a CPU 10, a memory 12, an HDD (hard disk drive) 14, a communication IF (interface) 16 for communicating data via a network (not shown), an output device 18 such as a display, It has an input device 20 including a pointing device such as a keyboard and a mouse, and an audio input device 22 such as a microphone that inputs audio and outputs an audio signal. These components are connected to each other through the bus 24 and input / output data to / from each other.

  FIG. 2 is a block diagram showing a functional configuration of the speech recognition system 1 according to the first embodiment of the present invention.

  As shown in FIG. 2, the voice recognition system 1 includes a voice input unit 100, a speech timing instruction acquisition unit 102, a voice signal holding unit 104, a speech segment detection unit 106, an erroneous operation detection unit 108, a voice recognition unit 110, a voice recognition dictionary. 112, an acoustic model 114, and an erroneous operation notification means 116. The function of the speech recognition system 1 is realized by loading a program into the memory 12 (FIG. 1) and executing it by the CPU 10. Note that all or part of the functions of the speech recognition system 1 may be realized by hardware.

  In the speech recognition system 1, the speech input means 100 inputs the speech signal output from the speech input device 22 (FIG. 1) and performs processing such as AD conversion and decoding of the encoded signal as necessary. Output digital signal of voice waveform.

  The utterance timing instruction acquisition unit 102 acquires an utterance timing instruction by the user including an utterance start instruction through the input device 20 (FIG. 1). The utterance timing includes at least the utterance start timing, and may include the utterance end timing. The instruction by the user is performed by pressing the button only before starting the utterance, uttering while pressing the button, releasing the button after the end of the utterance, pressing the button before starting the utterance and after the end of the utterance, etc. The utterance timing instruction acquisition means 102 acquires only the utterance start instruction or the utterance start and utterance end instructions by the operation method. When the utterance timing instruction acquisition unit 102 acquires the instruction, the utterance timing instruction acquisition unit 102, the voice signal holding unit 104, the erroneous operation detection unit 108, and the voice immediately after giving the acquired utterance timing instruction or given timing information of the instruction. Output to the recognition means 110.

  When the utterance timing instruction acquisition unit 102 acquires only the utterance start instruction, the utterance end timing may be substituted with the utterance end time detected by the utterance section detection unit 106 described later. Alternatively, the utterance end timing may be substituted by the utterance end time detected by the voice recognition unit 110 described later.

  The audio signal holding means 104 has a buffer for holding an audio signal for a predetermined time. The audio signal holding unit 104 stores the audio signal input from the audio input unit 100 in a buffer. When the capacity of the audio signal exceeds the capacity of the buffer, the audio signal holding means 104 may discard the oldest one in order and hold the audio signal for the latest predetermined time. Here, it is desirable that the buffer capacity is large enough to store a speech signal from the actual speech start to the speech start instruction when there is a delay in the speech start instruction. When a speech start instruction is input from the speech timing instruction acquisition unit 102, the speech signal holding unit 104 converts the speech signal stored in the buffer at that time together with the time information into the speech segment detection unit 106 and the speech recognition unit. To 110. The voice signal holding means 104 outputs the voice signal input from the voice input means 100 after the input of the utterance start instruction to the utterance section detection means 106 and the voice recognition means 110 together with the time information. When an utterance end instruction is input from the utterance timing instruction acquisition unit 102, the audio signal holding unit 104 stops outputting the audio signal and stores the input audio signal thereafter in the buffer.

  The utterance section detection unit 106 detects the user's utterance section from the voice signal output by the voice signal holding unit 104. The utterance section detecting means 106 detects the utterance section using, for example, a method based on the power information of the audio signal and the zero crossing. The utterance section detecting means 106 outputs information on the start time and end time of the detected utterance section.

  The erroneous operation detection means 108 detects an erroneous operation of the user's utterance timing instruction based on the utterance section detected by the utterance section detection means 106 and the utterance timing instruction acquired by the utterance timing instruction acquisition means 102. Specifically, the erroneous operation detection unit 108 includes information on start / end times of the utterance period input by the utterance period detection unit 106, presence / absence of an utterance timing instruction input by the utterance timing instruction acquisition unit 102, and time information. And an erroneous operation of the user's utterance timing instruction is detected.

  A method by which the erroneous operation detection unit 108 determines the presence / absence / type of an erroneous operation will be described later.

  The voice recognition unit 110 performs voice recognition on at least a part of the voice signal input by the voice signal holding unit 104. The voice recognition unit 110 performs voice recognition using the voice recognition dictionary 112, the acoustic model 114, and the like. The voice recognition unit 110 performs voice recognition by applying a technique using a hidden Markov model, for example. The voice recognition unit 110 outputs a text or a command as a recognition result.

  The speech recognition dictionary 112 stores a word set to be recognized and information on reading of each word.

  The acoustic model 114 stores an acoustic model obtained by modeling an acoustic pattern corresponding to reading.

  Note that the speech recognition unit 110 may determine a section to be recognized in the speech signal based on the time information of the speech timing instruction input from the speech timing instruction acquisition unit 102. For example, when an instruction to start speech and an instruction to end speech are input, the speech recognition unit 110 may limit the section to be recognized from the time of the speech start instruction to the time of the speech end instruction. Alternatively, the voice recognition means 110 has a buffer for holding a voice signal therein, and adds a certain margin to each of the utterance start time and the utterance end time, and gives an utterance end instruction from a predetermined time before the utterance start instruction. It may be limited to a certain time later. In addition, the voice recognition unit 110 may receive information on the presence or absence of an erroneous operation from the erroneous operation detection unit 108, and may stop the voice recognition process and the recognition result output for the recognition target section when there is an erroneous operation.

  When an erroneous operation is detected by the erroneous operation detection unit 108, the erroneous operation notification unit 116 notifies the user of a message corresponding to the type of the erroneous operation on a screen display or voice.

  Instead of this configuration, the voice recognition unit 110 includes the erroneous operation notification unit 116 so that the voice recognition unit 110 outputs a message or a recognition result corresponding to the type of the erroneous operation depending on whether there is an erroneous operation. May be.

  Next, a method for determining the presence / absence and type of an erroneous operation in the erroneous operation detection means 108 will be described.

  FIG. 3 is an illustration of a method for determining the presence / absence / type of an erroneous operation in the erroneous operation detection means 108.

  In FIG. 3, the utterance start indicates the start time of the utterance section detected by the utterance section detecting means 106. The utterance end indicates the end time of the utterance section detected by the utterance section detecting means 106. The utterance start instruction indicates the time of the utterance start instruction by the user acquired by the utterance timing instruction acquisition unit 102. The utterance end instruction indicates the utterance end instruction time acquired by the utterance timing instruction acquisition unit 102 by the user.

  The erroneous operation detection means 108 compares the time of the utterance start and the utterance start instruction. Further, the erroneous operation detection means 108 compares the time of the utterance end and the utterance end instruction. Next, the erroneous operation detection means 108 sequentially checks whether or not the comparison result matches each condition shown in FIG. 3, and if it matches any of the conditions, the presence / absence and type of the erroneous operation are determined from the corresponding determination result. judge. In FIG. 3, if it matches at least one of the conditions A to C, the erroneous operation detection means 108 determines that there is an erroneous operation of the type shown in the determination result column. If the condition D is met, the erroneous operation detection means 108 determines that there has been no erroneous operation.

  Specifically, the erroneous operation detection means 108 determines that the utterance start instruction is late when there is an utterance start instruction after the start of utterance as a result of the comparison. In addition, the erroneous operation detection means 108 determines that the utterance end instruction is early when there is an utterance end instruction before the end of the utterance as a result of the comparison. Further, as a result of the comparison, if there is no utterance end instruction within a predetermined time after the end of the utterance, the erroneous operation detection means 108 determines that the utterance end instruction has been forgotten. Further, as a result of the comparison, the erroneous operation detection means 108 determines that there is no erroneous operation when there is an utterance start instruction before the start of utterance and there is an utterance end instruction within a certain time after the end of the utterance.

  The erroneous operation notifying means 116 notifies the user of a message corresponding to the type of erroneous operation shown in FIG. 3 when there is an erroneous operation. For example, when it is determined that “the utterance start instruction is late”, the erroneous operation notifying unit 116 notifies the user of a message that prompts the user to speak after giving an instruction of the utterance start timing.

  Next, the operation of the voice recognition system 1 will be described.

  FIG. 4 is a flowchart showing the operation of the speech recognition system 1.

  As shown in FIG. 4, in step 10 (S10), the voice input means 100 performs a process such as decoding on the input voice signal and outputs the voice signal. Specifically, the voice input unit 100 inputs a voice signal from a microphone, performs AD conversion, and outputs a digital signal having a voice waveform.

  In step 12 (S12), the audio signal holding unit 104 stores the audio signal input from the audio input unit 100 in a buffer.

  In step 14 (S14), the utterance timing instruction acquisition means 102 determines whether or not an utterance timing instruction from the user has been accepted. If accepted, the utterance timing instruction acquisition means 102 sends the utterance timing instruction to the voice signal holding means 104, erroneous operation detection means. 108 and the voice recognition means 110, and the process proceeds to S16. Otherwise, the process returns to S12. For example, the utterance timing instruction acquisition unit 102 monitors the user's button pressing state, and detects an utterance start timing instruction when the button is pressed and an utterance end timing instruction when the button is released. The utterance timing instruction acquisition unit 102 outputs the detected instruction to the voice signal holding unit 104, the erroneous operation detection unit 108, and the voice recognition unit 110.

  In step 16 (S16), the voice signal holding means 104, when the voice start timing instruction is inputted from the voice timing instruction acquiring means 102, the voice signal stored in the buffer at that time together with the time information is spoken. It outputs to the zone detection means 106 and the voice recognition means 110.

  In step 18 (S18), the voice signal holding means 104 sends the voice signal input from the voice input means 100 after the notification of the utterance start timing instruction to the utterance section detection means 106 and the voice recognition means 110 together with the time information. Output.

  In step 20 (S 20), the utterance section detection unit 106 detects the utterance section from the voice signal output from the voice signal holding unit 104, and outputs the time information to the erroneous operation detection unit 108. Specifically, the utterance section detection unit 106 sequentially processes the audio signal output from the audio signal holding unit 104 and detects the start and end of the utterance using the calculated power information and the like.

  In step 22 (S22), the erroneous operation detection means 108 determines the user's utterance timing instruction based on the utterance section detected by the utterance section detection means 106 and the utterance timing instruction acquired by the utterance timing instruction acquisition means 102. Detecting misoperations. For example, the erroneous operation detection unit 108 compares the time information of the utterance start / utterance end input from the utterance section detection unit 106 with the presence / absence of the utterance timing instruction notified from the utterance timing instruction acquisition unit 102 and the time information. The erroneous operation detection means 108 determines the presence and type of a user's erroneous operation according to the determination criteria of FIG.

  If it is determined that there is no erroneous operation, in step 24 (S24), the speech recognition unit 110 recognizes the speech signal output from the speech signal holding unit 104 and outputs a recognition result.

  If it is determined that there is an erroneous operation, in step 26 (S26), the erroneous operation notification means 116 notifies the user of a message corresponding to the type of erroneous operation. For example, the erroneous operation notification means 116 outputs a message such as “Please speak after pressing the button” or “Please release the button after the utterance is finished” based on the conditions shown in FIG.

  Here, for the sake of simplicity, it has been described that the speech recognition means 110 performs speech recognition when it is determined in step 22 (S22) that there is no erroneous operation. Actually, the voice recognition unit 110 may start voice recognition at the time when an instruction of the utterance start timing is acquired, and may sequentially receive voice signals that are input to advance voice recognition. In this case, the voice recognition unit 110 may stop the voice recognition when it is determined that there is an erroneous operation.

  As described above, since the speech recognition system 1 according to the present embodiment does not perform the speech segment detection process until the notification of the speech start instruction is given, the processing load of the speech segment detection can be reduced. Thereby, since the speech recognition system 1 can reduce the influence of the utterance section detection processing load on the process of detecting an erroneous operation, it can accurately detect an erroneous operation of the user.

  Moreover, since the speech recognition system 1 limits the recognition target section based on the time information of the utterance timing instruction, the processing load for speech recognition can be reduced compared to the case where the recognition process is always performed.

  Furthermore, since the speech recognition system 1 includes the speech signal holding unit 104 and performs the speech segment detection process after a certain time from the speech start instruction, even if the speech start instruction is delayed from the actual speech start, The section can be detected with high accuracy.

<Second Embodiment>
Next, a second embodiment of the speech recognition system according to the present invention will be described.

  FIG. 5 is a block diagram showing a functional configuration of the speech recognition system 2 according to the second exemplary embodiment of the present invention.

  As shown in FIG. 5, the speech recognition system 2 according to the second exemplary embodiment of the present invention is compared with the speech recognition system 1 according to the first exemplary embodiment. The difference is that 106 outputs a voice signal to the voice recognition means 110. Another difference is that the utterance timing instruction acquisition unit 102 does not notify the voice recognition unit 110 of an utterance timing instruction.

  The utterance section detection unit 106 detects the user's utterance section from the voice signal input from the voice signal holding unit 104, and outputs information on the start / end time to the erroneous operation notification unit 116. At this time, the utterance section detecting means 106 may add a fixed-length margin before and after the utterance section. Further, the utterance section detecting unit 106 outputs the voice signal input from the voice signal holding unit 104 to the voice recognition unit 110.

  The voice recognition unit 110 performs voice recognition on a partial section of the voice signal input by the voice signal holding unit 104. The voice recognition unit 110 determines a section for speech recognition based on the utterance section detected by the utterance section detection unit 106.

  Other operations are the same as those in the first embodiment of the present invention.

  As described above, since the speech recognition system 2 according to the present embodiment limits the target of speech recognition based on the utterance section, the processing load can be reduced. This is because when there is no erroneous operation, the utterance section detected by the utterance section detection means 106 is a part of the section from the instruction to start utterance to the instruction to end utterance, and is shorter than the section in which the utterance timing is instructed. Because.

<Third Embodiment>
Next, a third embodiment of the speech recognition system according to the present invention will be described.

  FIG. 6 is a block diagram showing a functional configuration of the speech recognition system 3 according to the third exemplary embodiment of the present invention.

  As shown in FIG. 6, in the speech recognition system 3 according to the third exemplary embodiment of the present invention, the speech recognition unit 110 has the speech segment specifying unit 118 compared to the speech recognition system 2 according to the second exemplary embodiment. However, the point that the specified speech section information is output to the erroneous operation detection means 108 is different.

  In the present embodiment, for the sake of simplicity, discrete word recognition will be described as an example, but the present invention can be similarly applied to continuous word recognition.

  The voice recognition unit 110 performs voice recognition on a target section using the voice recognition dictionary 112 in which words to be recognized are stored.

  Specifically, the speech recognition unit 110 generates a standard pattern for each word using the acoustic model 114 based on the reading information of each recognition target word stored in the speech recognition dictionary 112. For example, when a phoneme HMM (Hidden Markov Model) is used as the acoustic model, the speech recognition means 110 connects the phoneme HMMs according to the reading of the word to form a standard pattern of the word. At that time, the voice recognition unit 110 adds a silent HMM to the front and back. The silent HMM is stored in advance in the acoustic model 114 as a model expressing background noise and the like. The voice recognition unit 110 compares the input voice signal cut out by the utterance section detection unit 106 with the standard pattern of each word, and calculates the likelihood for each word. The speech recognition unit 110 obtains the word with the highest likelihood and uses it as the recognition result.

  The utterance section specifying unit 118 specifies a section in which a word to be recognized is uttered among the sections to be subjected to speech recognition.

  Specifically, the utterance section specifying unit 118 associates the time between the input voice signal and the standard pattern of the word as the recognition result of the voice recognition unit 110. The utterance section specifying means 118 obtains a section that is associated with a portion of the input speech signal excluding the silence pattern before and after the word. The utterance section specifying unit 118 outputs start and end time information of the associated section to the erroneous operation detection unit 108.

  The voice recognition unit 110 also has a rejection function. Specifically, when the speech recognition unit 110 determines that the input speech signal does not match any of the recognition target words stored in the speech recognition dictionary 112, the speech recognition unit 110 rejects the recognition result.

  When the recognition result is rejected, the utterance section specifying unit 118 outputs information indicating that there is no utterance section to the erroneous operation detection unit 108.

  The erroneous operation detection unit 108 changes (for example, replaces) the speech segment detected by the speech segment detection unit 106 based on the segment specified by the speech segment specifying unit 118, and then erroneously operates the user's speech timing instruction. The presence / absence and type are determined. Note that the erroneous operation detection unit 108 may use the result of the utterance section specifying unit 118 without receiving the detection result of the utterance section detection unit 106.

  As described above, the speech recognition system 3 according to the present embodiment can distinguish between an actual speech segment and a noise segment in detail by performing speech recognition using information on a recognition target word. Therefore, the voice recognition system 3 can determine an erroneous operation using information on an actual utterance section, that is, a more accurate utterance section.

  In the voice recognition system 3, since the voice recognition unit 110 has a rejection function and can cancel a section that is not a user's speech intended for voice input, the speech section can be detected with high accuracy. Therefore, the voice recognition system 3 can accurately detect a user's erroneous operation.

<Fourth Embodiment>
Next, a fourth embodiment of the speech recognition system according to the present invention will be described.

  FIG. 7 is a block diagram showing a functional configuration of the speech recognition system 4 according to the fourth embodiment of the present invention.

  The utterance timing instruction acquisition unit 102 outputs an utterance timing instruction by the user including an utterance start instruction to the audio signal holding unit 104 and the erroneous operation detection unit 108.

  The voice signal holding unit 104 holds the input voice signal, and outputs the held voice signal to the utterance section detection unit 106 when an utterance start instruction is input by the utterance timing instruction acquisition unit 102. Further, the voice signal holding means 104 outputs a voice signal inputted thereafter to the utterance section detecting means 106 when the utterance start timing instruction is inputted by the utterance timing instruction obtaining means 102.

  The utterance section detecting unit 106 detects the utterance section from the voice signal output from the voice signal holding unit 104.

  The erroneous operation detection unit 108 detects a user's erroneous operation based on the speech segment detected by the speech segment detection unit and the speech timing instruction acquired by the speech timing instruction acquisition unit 102.

  As described above, according to the speech recognition system 4 according to the present embodiment, it is possible to reduce the processing load for detecting the utterance section and accurately detect a user's erroneous operation.

  The voice recognition system according to the present invention is applicable to uses such as a voice recognition apparatus that performs voice data input, text input, and device operation instructions.

1 Voice recognition system 1
2 Speech recognition system 2
3 Voice recognition system 3
4 Voice recognition system 4
10 CPU
12 memory 14 HDD
16 Communication IF
18 output device 20 input device 22 voice input device 24 bus 100 voice input means 102 utterance timing instruction acquisition means 104 voice signal holding means 106 utterance section detection means 108 erroneous operation detection means 110 voice recognition means 112 voice recognition dictionary 114 acoustic model 116 notification of erroneous operation Means 118: Speaking section specifying means

Claims (10)

An utterance timing instruction acquisition means for acquiring an utterance timing instruction by the user including an utterance start instruction;
An audio signal holding unit that holds an input audio signal and outputs an audio signal that is input after that when the instruction to start utterance is acquired by the utterance timing instruction acquisition unit; and
An utterance section detecting means for detecting an utterance section from the voice signal output by the voice signal holding means;
Compare the time information of the utterance section detected by the utterance section detection means with the presence / absence and time information of the utterance timing instruction acquired by the utterance timing instruction acquisition means, and at least the time of the instruction to start the utterance Erroneous operation detection means for detecting as an erroneous operation of the user when is later than the start time of the utterance section ,
A speech recognition system comprising:
The utterance timing instruction further includes an utterance end instruction, and the erroneous operation detection unit further includes the utterance end instruction when the utterance end instruction time is earlier than the end time of the utterance section or when there is no utterance end instruction itself. Detect as user's misoperation

The speech recognition system according to claim 1.
The audio signal holding means holds the latest predetermined time of the input audio signal.
The speech recognition system according to claim 1 or 2 . Voice recognition means for performing voice recognition on at least a part of the voice signal output by the voice signal holding means;
Speech recognition system according to any of claims 1-3. The voice recognition means determines a section for voice recognition based on the utterance section detected by the utterance section detection means;
The voice recognition system according to claim 4 . The voice recognition means stops voice recognition when the erroneous operation detection means detects an erroneous operation;
The speech recognition system according to claim 4 or 5 . The speech recognition means performs speech recognition using a speech recognition dictionary in which a recognition target word is stored, and identifies a section in which the recognition target word is uttered among the partial sections.
The erroneous operation detection means detects an erroneous operation of the user after changing the utterance section detected by the utterance section detection means based on the section specified by the voice recognition means.
The speech recognition system according to any one of claims 4 to 6 . When the erroneous operation detection means detects an erroneous operation, it further includes an erroneous operation notification means for notifying a message according to the type of the detected erroneous operation.
Speech recognition system according to any of claims 1-7. Get the utterance timing instruction by the user including the utterance start instruction,
When an input voice signal is held and the instruction to start speech is acquired, the held voice signal and a voice signal input thereafter are output.
Detecting an utterance section from the output voice signal,
Comparing the time information of the detected utterance section with the presence / absence and time information of the acquired utterance timing instruction, at least when the time of the utterance start instruction is later than the start time of the utterance section Detect as user's mistaken operation,
Speech recognition method. On the computer,
An utterance timing instruction acquisition step for acquiring an utterance timing instruction by the user including an utterance start instruction;
An audio signal holding step for holding an input audio signal and outputting an audio signal input after that when the instruction to start utterance is acquired by the utterance timing instruction acquisition step;
An utterance interval detection step of detecting an utterance interval from the audio signal output by the audio signal holding step;
The time information of the utterance section detected by the utterance section detection step is compared with the presence / absence and time information of the utterance timing instruction acquired by the utterance timing instruction acquisition step, and at least the instruction start instruction time is An erroneous operation detection step of detecting as an erroneous operation of the user when it is later than the start time of the utterance section ,
Voice recognition program that executes

Images