JP2009069202A - Speech processor - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To easily and surely cancel speech recognition by an operator when erroneous recognition of the speech recognition occurs. <P>SOLUTION: The operator inputs speech from a microphone to operate a sound device 14. When speech recognition results of a speech recognition unit/device control unit 12 are erroneous, the operator performs specific movement such as hands are shaken right and left. A moving body detecting unit 18 detects movement of the hands of the operator by non-contact, and returns to a just-before operation state when detecting the specific hand movement. When the hands are shaken from left to right, the operation may be cancelled, and when the hands are shaken from right to left, it may start speech recognition. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to a voice processing device, and more particularly to cancellation of a voice recognition result.

  A technique for controlling the operation of a device by an operator's utterance is known, but when the device is a device that outputs audio, such as a DVD drive, a car audio device, or a car navigation system, the operator's utterance and the device Are likely to cause misrecognition. For this reason, it is desired that the operator can easily and reliably cancel the recognition result when erroneous recognition occurs.

  Patent Document 1 below discloses that a speech recognition result is invalidated when an invalid instruction is input before a predetermined time has elapsed since the speech recognition. Examples of input from a keyboard or a mouse are illustrated.

  Patent Document 2 below also discloses that the immediately preceding voice command is invalidated by voice input such as “Chiga”, “Trikes”, and “Muko”.

Japanese Patent Application Laid-Open No. 7-219583 Japanese Patent Laid-Open No. 7-219591

  However, if the operator's utterance and the voice from the device are mixed, it is likely that misrecognition is likely to occur. When canceling or invalidating with the operator's voice such as “Chigau” or “Trikes”, the cancellation or invalidation Therefore, the utterance itself may be misrecognized. On the other hand, in the method of canceling or invalidating by inputting a command from a keyboard, mouse, etc., the operator is forced to operate these input devices, so the advantage of operating the device by voice recognition is only lost. In addition, there is a problem that cannot be canceled if the operator cannot operate these input devices. Of course, even if the voice recognition itself is correct, the operator may change his / her intention. In this case, it is desirable that it can be canceled quickly.

  An object of the present invention is to provide an apparatus capable of canceling (or invalidating) a recognition result easily and surely when erroneous recognition of voice recognition occurs.

  The present invention is a voice processing device that recognizes an operator's voice and executes processing including voice output processing, wherein the voice recognition means for recognizing the operator's voice and the movement of the operator's hand A moving object detecting means for detecting by contact; and a control means for canceling the immediately preceding voice recognition result by the voice recognizing means when the first movement of the operator's hand is detected by the moving object detecting means after the voice recognition. It is characterized by.

  In one embodiment of the present invention, the control means starts voice recognition by the voice recognition means when the moving body detection means detects a second movement of the operator's hand.

  In another embodiment of the present invention, the control unit may display a voice recognition result immediately before by the voice recognition unit when the moving body detection unit detects a third movement of the operator's hand after voice recognition. Determine.

  According to the present invention, when erroneous voice recognition or the like occurs, it can be easily and quickly canceled.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings.

  FIG. 1 is a block diagram showing the configuration of the speech processing apparatus according to this embodiment. The microphone 10 receives an utterance from an operator (user), converts it into an electrical signal, and outputs it to the voice recognition unit / device control unit 12.

  The speech recognition unit / device control unit 12 analyzes the speech of the operator from the microphone 10 and recognizes the speech, and outputs a control command to the acoustic device 14 according to the recognition result. A known technique can be used for speech recognition. A speech database may be stored in advance, and speech recognition may be performed by collating the input utterance with the speech database.

  The acoustic device 14 is a variety of devices such as a DVD player, car audio, car navigation, game machine, and communication terminal device, and is a device that outputs at least sound. Control commands from the voice recognition unit / device control unit 12 are, for example, a playback command, a stop command, a fast-forward command, a recording / recording command, a scroll command, a tray open / close command, and the like. The audio signal from the acoustic device 14 is supplied to the speaker 16 and output as audio.

  The moving body detection unit 18 detects the movement of the operator's hand in a non-contact manner, and outputs different detection signals to the voice recognition unit / device control unit 12 and the operation memory unit 20 according to the movement of the hand. The moving body detection unit 18 detects the movement of the hand using, for example, an infrared detector. The movement of the operator's hand is a movement of the hand that can be easily realized by the operator, such as opening and closing of the hand and swinging of the hand from side to side. In the present embodiment, a case where a left / right swing of the hand is detected is illustrated. The moving body detection unit 18 detects the left / right swing of the hand, and detects the case where the hand is swung from left to right and the case where the hand is swung from right to left. Then, the case where the hand is shaken from the left to the right is output to the voice recognition unit / device control unit 12 as a trigger signal for starting the speech recognition, and the case where the hand is shaken from the right to the left is canceled (UNDO). A signal (or a return signal) is output to the operation memory unit 20.

  The operation memory unit 20 stores the operation state of the acoustic device 14 before performing a new operation according to the voice recognition result. When a cancel signal is supplied from the moving object detection unit 18, the operation memory unit 20 outputs the previous operation state to the voice recognition unit / device control unit 12 and outputs a current command, that is, a command corresponding to the voice recognition result. Cancel and return to the previous operating state. For example, when a stop command is output as a result of voice recognition while the audio device 14 is being played back, the operation memory unit 20 supplies the playback state to the voice recognition unit / device control unit 12 as the previous operation state. In response to this, the voice recognition unit / device control unit 12 outputs a reproduction command to the acoustic device 14.

  FIG. 2 shows the configuration of the moving object detection unit 18. The moving body detection unit 18 includes two infrared detectors 30 and 32 and a direction detector 34. The two infrared detectors 30 are arranged close to each other by a predetermined distance in the left-right direction, and each detects infrared rays and outputs them to the direction detector 34.

  The direction detector 34 includes two comparators 36 and 38 and two flip-flops 40 and 42. The infrared detector 30 is connected to one input terminal (+) of the comparator 36, and a predetermined voltage (threshold voltage) is applied to the other input terminal (−). The comparator 36 compares the output of the infrared detector 30 with a predetermined threshold voltage, and outputs a Hi or Low binary signal to the flip-flops 40 and 42 according to the magnitude relationship. The infrared detector 32 is connected to one input terminal (+) of the comparator 38, and a predetermined voltage (threshold voltage) is applied to the other input terminal (−). The comparator 38 compares the output of the infrared detector 32 with a predetermined threshold voltage, and outputs a Hi or Low binary signal to the flip-flops 40 and 42 according to the magnitude relationship.

  The flip-flops 40 and 42 are D-type flip-flops. The output of the comparator 38 is supplied to the D terminal of the flip-flop 40, and the output of the comparator 36 is supplied to the clock (CK) terminal. The output of the comparator 36 is supplied to the D terminal of the flip-flop 42, and the output of the comparator 38 is supplied to the clock (CK) terminal. Accordingly, the Hi output of the comparator 38 is output at the timing of the Hi output of the comparator 36, and the Hi output of the comparator 36 is output at the timing of the Hi output of the comparator 38. The comparator 36 outputs Hi when the infrared detector 30 detects infrared rays emitted from the operator's hand, and the comparator 38 outputs Hi when the infrared detector 32 detects infrared rays emitted from the operator's hand. Eventually, as indicated by A in the figure, when the operator first places his hand in front of the infrared detector 30 and then shakes his hand in front of the infrared detector 32, the flip-flop 42 outputs Hi. Is supplied to the operation memory unit 20, and the operator first places his hand in front of the infrared detector 32 and then shakes his hand in front of the infrared detector 30 as shown in FIG. The Hi output from the flip-flop 40 is supplied to the device controller 12. If the Hi output of the flip-flop 40 is used as a trigger signal for voice recognition and the Hi output of the flip-flop 42 is used as a cancel signal, the voice recognition is canceled when the hand is waved in the A direction and the hand is waved in the B direction. The trigger / decision can be indicated. If the direction A in the figure corresponds to the direction of the hand swing from right to left, and the direction B in the figure corresponds to the direction of hand swing from left to right, the operator can start voice recognition by simply changing the direction of hand swing. It is possible to instruct differently from cancellation.

  FIG. 3 shows an overall process flowchart of the present embodiment. When the device is activated, the voice recognition unit / device control unit 12 determines whether or not the operator has shaken his / her hand from left to right (S101). Specifically, it is determined whether a detection signal is received from the moving object detection unit 18 when the hand is swung from left to right. When the detection signal when the hand is shaken from the left to the right is received, the voice recognition unit / device control unit 12 starts a predetermined voice recognition process (S102), and analyzes the voice of the operator input from the microphone 10 (S103, S104). If no detection signal is received when the hand is swung from left to right, speech recognition does not start.

  When the voice recognition is started, the operation memory unit 20 stores the operation state before starting the voice recognition or when the voice recognition is started (these are collectively referred to as a previous state) (S105). For example, if the voice recognition is started, the stop state is stored if the voice recognition is started. If it is in a playback state when voice recognition is started, the playback state is stored. The playback position may be further stored. The storage of the current state is known as a resume function in a computer or the like. After storing the previous operation state, the voice recognition unit / device control unit 12 outputs a control command corresponding to the voice recognition result to the acoustic device 14, and the acoustic device 13 performs an operation corresponding to the command (S106). .

  After the acoustic device 14 performs an operation according to the command, it is determined whether or not the operator has shaken his / her hand from right to left (S107). Specifically, it is determined whether or not a detection signal is received from the moving object detection unit 18 when the hand is swung from right to left. When the detection signal when the hand is swung from right to left is received, the voice recognition unit / device control unit 12 cancels the command executed in S106, reads the immediately previous operation state stored in the operation memory unit 20, and performs sound processing. The device 14 is returned to the previous operating state (S108). On the other hand, if the detection signal when the hand is swung from right to left is not received, the command executed in S106 is continuously executed.

  Here, the determination in S107 may be performed within a predetermined time after the execution of a new command as a result of speech recognition in S106. That is, it is determined whether or not the operator swings his / her hand from right to left within a predetermined time after starting execution of a new command. The command is canceled only when the operator shakes his / her hand from right to left within a predetermined time, and when the predetermined time elapses, the operator continues to execute the command assuming that there is no intention to cancel. Or when it determines with NO by S107, you may determine whether the operator further swung the hand from the left to the right. Specifically, it is determined whether or not a detection signal when the hand is shaken from the left to the right is received from the moving object detection unit 18 as in the determination of S101. Then, when the operator does not shake his hand from right to left, but from left to right, the command executed in S106 is confirmed and the command is continuously executed by assuming that the command executed in S106 is affirmed by the operator. . In this way, the operator simply cancels the voice recognition result when the hand shakes his hand from right to left and decides (or confirms) when his hand shakes from left to right. Just shake it. The advantage that it is possible to instruct cancellation / determination of the result of speech recognition simply by waving a hand becomes prominent when speech is output from the acoustic device 14 by a command corresponding to the result of speech recognition. In other words, it is difficult to instruct cancellation or determination by voice in a situation where sound is output from the acoustic device 14 because the possibility of misrecognition increases, but such a problem arises if the hand is shaken. Absent.

  In the present embodiment, the two infrared detectors 30 and 32 arranged close to each other on the left and right sides are used as the moving body detection unit 18, but it is needless to say that the present invention is not limited to this. Any detector that can detect the movement of the operator's hand in a non-contact manner is included, such as a light detector, an ultrasonic detector that detects ultrasonic waves, and a camera that detects images.

  In this embodiment, the hand movement of the operator is exemplified as the movement of the hand. However, opening and closing the hand, shaking the finger, moving the finger, making the finger a specific shape, and the like are also included. For example, it can be configured to cancel by waving an index finger and to instruct a decision by placing a thumb.

It is a configuration block diagram of an embodiment. FIG. 2 is a configuration block diagram of a moving object detection unit in FIG. 1. It is a processing flowchart of an embodiment.

Explanation of symbols

  10 microphone, 12 voice recognition unit / device control unit, 14 acoustic device, 16 speaker, 18 moving object detection unit, 20 operation memory unit.

Claims (3)

A voice processing device that recognizes an operator's voice and executes processing including voice output processing,
Voice recognition means for recognizing the voice of the operator;
Moving object detection means for detecting the movement of the operator's hand in a non-contact manner;
Control means for canceling the immediately preceding voice recognition result by the voice recognition means when the first movement of the operator's hand is detected by the moving object detection means after voice recognition;
A speech processing apparatus comprising: The apparatus of claim 1.
The voice processing apparatus, wherein the control means starts voice recognition by the voice recognition means when the moving body detection means detects a second movement of the operator's hand. The apparatus according to claim 1,
The speech processing apparatus characterized in that the control means determines a speech recognition result immediately before by the speech recognition means when a third movement of the operator's hand is detected by the moving object detection means after speech recognition.

Images