JP2009145434A - Speech recognition system - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a speech recognition system which represents a recording state, a state in progression of speech recognition and the completion state of speech recognition by utilizing waveform. <P>SOLUTION: The speech recognition system comprises at least one speech recognition engine and a display device, wherein a signal indicator interface and a letter output interface are disposed on the display device. Between both interfaces, the signal indicator interface represents a speech signal input by a user by waveform and displays the recording state, the state in progression of speech recognition or the completion state of speech recognition. The letter output interface is used for displaying a character result which is speech-recognized and the character result includes at least one word unit. Further, the displayed word unit and a corresponding waveform unit are connected to one set of feedback adjustment selection items respectively; the set of feedback adjustment selection items includes at least one feedback adjustment selection item, makes the user select a command to correct a speech recognition error or perform feedback adjustment of the speech recognition system. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to a speech recognition system, and in particular, applied to an electronic device of an image display interface such as a desktop personal computer, a notebook personal computer, a home multimedia system, a television, a DVD, a video system, a mobile phone or a PDA, and has a waveform. The recording status, speech processing status or speech recognition completion status can be displayed, and at least one feedback adjustment selection item is provided on the waveform and character display interface, and the user selects a command and corrects the speech recognition error. Or relates to a visual feedback interface of speech recognition technology for adjusting a speech recognition system.

  In recent years, the development of voice recognition technology has made the user more convenient in the usage of many electronic devices. Any desktop PC, notebook PC, mobile phone, PDA or other electronic device requires input and / or limb coordination to achieve control objectives. For example, when a user operates a personal computer, it needs the assistance of a keyboard, mouse or other attached control device to complete command input. Or, use a touch-type monitor to simplify program input. However, the monitor display area is limited, and it is necessary to perform operations using fingers, and the maximum convenience is still achieved. Absent. The above problem may only be inconvenient for the general public, but it is difficult to operate the electronic device for a physically handicapped person, a person with motor nerve disease, or a visually impaired user. It is. Speech recognition technology can exactly solve these problems.

  The application direction of voice recognition is that the user needs only a voice signal input device, for example, a microphone, and the voice recognition system recognizes the voice input, and then outputs a character corresponding to the voice to be input, or further Direct command operation is performed based on the acquired voice recognition result.

  When using the voice recognition system, as described above, the user inputs and records voice by the voice signal input device, and then starts the recognition process. During the recording and recognition process, many factors affect the final speech recognition result, such as the type of speech signal input device used, the recording environment, the distance from the speech input device, and the like. Therefore, there is a need for monitoring and control of recording and recognition processes. In this regard, there is a conventional technique or using a different image to display the recording state and the recognition state, or using a change in the image to display the recording or recognition state. However, simultaneously with displaying the status, it is impossible to reflect the superiority or inferior quality of the recording or recognition result or whether the recording process is successful.

Some of the speech recognition results provide some adjustment functions based on the prior art or the recognition results, but most of them are provided only for the entire speech recognition results. It is not possible to improve the function of the speech recognition system through feedback by adjusting the heel part, and it is not possible to meet the demand for each user. For example, if there is a special utterance in the pronunciation of words or words by user and if the voice recognition results for the characters and words cannot be fed back or adjusted, the voices tailored to each user The recognition system cannot be provided, and the effect of its actual operation is greatly reduced.
JP 2003-167600 A

  An object of the present invention is to provide a voice recognition system that uses a waveform to indicate a recording state, a voice recognition in progress state, or a voice recognition completion state. The quality of the voice recognition result is monitored and controlled.

  Another object of the present invention is to provide a speech recognition system with an error correction feedback adjustment mechanism to provide a user with efficient speech recognition error correction or feedback adjustment of the speech recognition system.

  In order to achieve the above object, a speech recognition system provided by the present invention includes at least one speech recognition engine and a display device, and includes a signal indicator interface and a character output interface on the display device, and the signal indicator includes The interface represents a voice signal input by the user with a waveform, and displays a recording state, a voice recognition in progress state, or a voice recognition completion state. The character output interface is used to display a voice-recognized character result, and the character result includes at least one word unit. In addition, the word unit to be displayed and the corresponding waveform unit are each connected with one feedback adjustment selection item set; the feedback adjustment selection item set includes at least one feedback adjustment selection item, and allows the user to select a command; Correct the speech recognition error or feedback adjust the speech recognition system.

  The waveform preferably represents the recording state, the state of speech recognition signal, and the state of speech recognition completion in different colors.

  After the input speech signal is recognized, the recognition result is output on the character output interface, and each word unit is preferably checked against the corresponding waveform unit on the signal indicator interface. The speech recognition quality is roughly divided into about three types, including a state in which strict inspection and correction are required due to good quality, poor quality, and poor quality, and are displayed in different colors.

  The waveform unit of the signal indicator interface and one feedback adjustment selection item set are connected; the feedback adjustment selection item set includes a connection of at least one selection item, and allows the user to select a command, thereby efficiently performing a voice recognition error. Or correct the feedback of the speech recognition system.

  The word unit of the character output interface is connected to one feedback adjustment selection item set. The feedback adjustment selection set includes a concatenation with at least one selection item to allow the user to select a command and efficiently correct a speech recognition error or feedback adjust the speech recognition system.

The voice recognition system that represents the user's voice signal using the waveform provided by the present invention allows the user to immediately determine whether the recording process is successful and the quality of the input voice signal.
The speech recognition system that changes the color of the waveform provided by the present invention allows the user to conveniently monitor and control the progress of speech processing and the quality of speech recognition results.
With the speech recognition system provided by the present invention, the user can conveniently complete the character input operation by correcting the error or adjusting the feedback of the system for the speech signal input by the word unit and the characters of the speech recognition result. Alternatively, the improvement of the performance of the voice recognition system can be continued.

A preferred embodiment of the voice recognition system of the present invention will be described with reference to the drawings.
FIG. 1 is an explanatory diagram of a voice recognition system using a visual feedback interface of a voice recognition technique according to an embodiment of the present invention. A desktop personal computer, a notebook personal computer, a home multimedia system, a television, a DVD, and an image. The present invention relates to a speech recognition system applied to an electronic device of an image display interface such as a system, a mobile phone or a PDA.
As shown in the figure, the speech recognition system according to the embodiment of the present invention includes at least a speech recognition engine 10 and a display device 20, and the display device 20 is provided with a signal indicator interface 30 and a character output interface 40. Among them, the signal indicator interface 30 represents a voice signal input by the user by a waveform 32 and displays a voice recognition state of a recording state. The character output interface 40 displays a speech recognition result character 42, the character result including at least one word unit. As shown in the figure, the waveform 32 displayed on the signal indicator interface 30 is used to display a voice signal input by the user, and the character 42 displayed on the character output interface 40 is acquired after the voice recognition. It is a result.

  The display device 20 of the speech recognition system according to the embodiment of the present invention is used as a display screen of an electronic device such as a desktop personal computer, a notebook personal computer, a home multimedia system, a television, a DVD, an AV device, a mobile phone, or a PDA. It can be used and can be connected to a display screen for outputting a video signal or a display screen on a remote control.

FIG. 2 is an explanatory diagram of the first embodiment of the present invention, and displays the recording state of the user. As shown in the figure, during the recording process, the user inputs a voice into the voice recognition system by a voice signal input device (not shown, for example, a microphone), and the input voice signal has a waveform 32 and a signal indicator interface 30. Displayed above. The use of the waveform has two advantages. One of the advantages is that if the user cannot input the user's voice signal smoothly for a certain reason during the recording process, for example, the voice signal input device does not start. Causes such as poor contact between the electronic device provided with the voice input device and the voice recognition system, and at this time, the user can observe the change in the waveform and react immediately, thereby avoiding wasting time.
The other is that the user can roughly determine the input quality of the audio signal based on the waveform shape, and can make an appropriate adjustment. For example, environmental noise interference, the sensitivity of the audio input device used, or the method in which the user uses the audio input device may affect the quality of the audio signal input. Grasping and eliminating the factors that affect signal input quality is an auxiliary that cannot be neglected in the subsequent speech recognition process.

  As described above, the signal indicator interface 30 of the first embodiment of the present invention displays the recording state and the voice recognition state by the waveform 32, and the voice recognition state further includes the voice recognition in progress state and the voice recognition completion state. including. In addition, representatives representing the recording state, the voice recognition in progress state, and the voice recognition completion state are displayed in different colors, and the user can visually determine the processing state, the voice recognition quality, or the speed of voice recognition.

  When the voice signal input by the user is in the process of identification, the signal waveform 32 of the signal indicator interface 30 displays the signal waveform already processed in different colors, indicating the process of voice recognition. In other words, at the start, the audio signal input by the user is displayed in the recording state color. After the voice recognition process is started, the processed voice signal is displayed in the color of the voice recognition in progress state. After all of the input speech signal recognition is completed, it is displayed in the color of the speech recognition completion state. The 某 word unit is a color with good recognition quality, the 某 word unit is a color with poor recognition quality, and the 某 word unit is a color with poor recognition quality.

  As shown in FIG. 3, a solid line waveform 321 is a completed speech recognition process, and a dotted line waveform 322 is a completed speech recognition process. After the voice recognition is completed, all the waveforms are changed to a new color to display that the voice recognition process is completed. Details will be described later.

  After the user inputs the voice signal and the input voice signal is also recognized, the recognition result is preferably selected on the word output interface 40 by selecting the word unit 420 one by one. As shown in FIG. 4, a voice signal input by the user is displayed as a waveform 32, of which the waveform 32 divides at least one waveform unit 320, and each waveform unit 320 is a word unit in the recognition result. Corresponding to 420, the correspondence between the two corresponds by a method of adjusting the front and rear positions to match each other. In this embodiment, each waveform unit 320 corresponds to the word unit 420 of the recognition result. As can be seen from FIG. 4, the user inputs a voice signal “how is the weather today”, and the character result displayed by the user is “how is the weather today”. The input speech signal corresponding to the waveform unit 320 may be the “today” word unit 420 in the recognition result. The positions before and after both are collated with each other, and the recognition quality is expressed by different colors.

  When the speech recognition system is used for speech understanding, the signal indicator interface 30 still outputs the result of speech understanding, as shown in FIG. In addition, the character output interface 40 may include the speech recognition result character 42, or it may be hidden first and displayed after the user has selected to display.

  As shown in FIG. 4, after the speech recognition is completed, the signal and character display are arranged in units of word units 420 such as “today”, “weather”, “what”, etc., and each word unit is Displayed in color to indicate the quality of the speech recognition result of the word unit. In this embodiment, each word unit 420 is displayed in green, yellow or red. Of these, green indicates that the character has good speech recognition quality. Yellow warns that the character has poor voice quality. Red indicates a speech recognition result where the character has poor speech quality and needs to be inspected and corrected. In this way, the user can visually distinguish the superiority or inferiority of the speech recognition result quality of each word unit, and can perform appropriate error correction and feedback adjustment system processing.

Each waveform unit 320 is connected to a feedback adjustment selection item set.
The selection set includes at least one selection item that causes the user to re-record, correct the recording, correct the speech recognition error, or feed back the speech recognition system. As shown in FIG. 5, each waveform unit 320 is connected to a first feedback adjustment selection item set 50, and the first feedback adjustment selection item set 50 includes at least one feedback adjustment selection item 52. In the present embodiment, the first feedback adjustment selection item set 50 is a first feedback adjustment selection such as “repeat”, “re-record”, “acceptance adjustment”, “change to handwritten input”, “change to keyboard input”, etc. Item 52 is included.
After the speech recognition is completed, the user moves the mouse pointer 22 displayed on the display device 20 to the waveform unit 320 to be feedback-adjusted, points automatically or with a mouse or a touch pad, and connects to the waveform unit 320. One feedback adjustment selection item set 50 is displayed on the display device 20. Accordingly, the user selects the necessary first feedback adjustment selection item 52, corrects the speech recognition result, or performs feedback adjustment of the speech recognition system.

For example, when an abnormality is found in the waveform, the user selects the selection item “repeat” of the first feedback adjustment selection item 52, repeats “the audio signal”, and determines whether there is noise interference. To do. Alternatively, when the character result of the voice recognition has a large deviation, the voice signal input by the first feedback adjustment selection item 52 set as the “repeat” is heard again, and the location of the cause, for example, the deviation of the pronunciation is detected. .
When it is determined that there is a problem, the selection item “re-record” of the first feedback adjustment selection item 52 is selected, and the audio signal is input again. If the deviation of the character output result of speech recognition is due to the problem of the original pronunciation habit, select the “acceptance adjustment” selection item of the first feedback adjustment selection item 52, sharply adjust the speech recognition system, and use the Can be adapted to the requirements of the person.
Before adjusting the speech recognition system so that the words can be clearly recognized, the user decides to change the input mode and selects, for example, “change to handwriting input” or “change to keyboard input” and the like. Switch the input mode to handwriting or keyboard mode and complete the input purpose.

Each word unit 420 in the character display result is also connected to the feedback selection item set. The selection item set includes at least a connection with a feedback adjustment selection item to allow the user to select, correct a speech recognition error, or feedback adjust the speech recognition system.
As shown in FIG. 6, each word unit 420 is connected to a second feedback adjustment selection item set 60, and the second feedback adjustment selection item set 60 includes at least one feedback selection item 62. In this embodiment, the second feedback adjustment selection item set 60 includes “display the next voice recognition selection item”, “display the voice recognition selection item preferentially according to the similarity of the voice”, “all selections”. Feedback selection items 62 such as “display item”, “change to handwritten input”, and “change to keyboard input” are included. After the speech recognition is completed, the user moves the mouse pointer 22 onto the word unit 420 to be feedback-adjusted and points to the word unit 420 automatically or by pointing with the mouse or the touch pad. The item set 60 can be displayed on the display device 20. Therefore, the user can perform feedback adjustment on the speech recognition result by selecting the necessary feedback adjustment selection item 62.

  Due to the problem of pronunciation, the voice recognition character result obtained based on the voice signal input by the user may change greatly. Using FIG. 6 as an example, the speech read by the user is “I want to eat rice”, and the result of recognition may differ depending on the pronunciation habit. Based on the audio signal input by the user, the visual feedback system of the present invention allows the user to select a plurality of speech recognition approximation results of the input audio signal. The approximation result can be determined by selecting a different feedback selection item 62 in the second feedback adjustment selection item set 60. For example, by selecting the “next option” selection item 62, the user can obtain the next option. By selecting the “voice similarity priority” selection item 62, the user can detect and obtain the most possible option based on the relationship between the preceding and following words. Alternatively, by selecting the selection item 62 “display all options”, the user can display all the voice recognition options. Alternatively, the user can select other input modes, for example, select “Switch to handwriting input” or “Switch to keyboard input”, switch the voice input mode to handwriting or keyboard input mode, and Can be completed.

  In the present invention, the preferred embodiments have been disclosed as described above, but these are not intended to limit the present invention in any way, and anyone who is familiar with the technology can make an equivalent scope without departing from the spirit and scope of the present invention. Of course, various fluctuations and hydration colors can be added.

It is explanatory drawing of the speech recognition system of this invention. It is explanatory drawing of 1st Example of the speech recognition system of this invention. It is another explanatory drawing of 1st Example of the speech recognition system of this invention. It is another explanatory drawing of 1st Example of the speech recognition system of this invention. It is a use state figure of the 1st example of the voice recognition system of the present invention. It is another use condition figure of 1st Example of the speech recognition system of this invention.

Explanation of symbols

10 voice recognition engine 20 display device 22 mouse pointer 30 signal indicator interface 32, 321, 322 waveform 320 waveform unit 40 character output interface 42 character 420 word unit 50 first feedback adjustment selection item set 52 first feedback adjustment selection item 60 second Feedback adjustment selection item set 62 Second feedback adjustment selection item

Claims (13)

Including at least one speech recognition engine and a display device;
And on the display device,
A signal indicator interface that represents the voice signal input by the user in the waveform and displays the recording status, voice recognition in progress status, voice recognition completion status,
A character output interface that is used to display a voice-recognized character result and the character result includes at least two word units;
A speech recognition system comprising:   2. The voice recognition system according to claim 1, wherein waveforms of a recording state, a voice recognition in progress state, and a voice recognition completion state displayed on the signal indicator interface are displayed in different colors.   2. A speech recognition system according to claim 1, wherein each word unit of the character result of speech recognition on the character output interface represents the speech recognition quality of each word unit with a different color.   Each word unit is displayed in green, yellow, or red, of which green indicates good speech recognition quality, yellow warns that it has poor speech recognition quality, red is poor speech recognition 4. The speech recognition system according to claim 3, wherein the speech recognition system has quality and indicates that strict inspection and correction are required.   Each of the word units is connected to a feedback adjustment selection item set, and the selection item set includes at least one feedback adjustment selection item, which allows the user to select and correct a speech recognition error or feedback a speech recognition system. 4. The speech recognition system according to claim 3, wherein adjustment is performed.   A mouse pointer to be displayed on the display device is moved to a word unit that the user wants to perform feedback adjustment, or is pointed with a touch pad or a mouse, and the feedback adjustment selection item group is displayed on the display device. The voice recognition system according to claim 5.   The feedback selection items included in the feedback adjustment selection item group connected to the word unit are “next option”, “display voice recognition options based on priority of voice similarity”, “display all approximate recognition results”, “handwritten input” The voice recognition system according to claim 5, wherein the voice recognition system is “change to”, “change to keyboard input”, or any combination thereof.   The speech recognition completion waveform on the signal indicator interface further includes at least one waveform unit, and each waveform unit corresponds to and contrasts with one word unit of the speech recognition result displayed by the character output interface. 5. The voice recognition system according to claim 3, wherein the voice recognition quality of the word units is displayed in the same color.   The waveform unit of the signal indicator interface is connected to a feedback adjustment selection set, and the selection set includes at least one feedback adjustment selection, which allows the user to make duplicate recordings, correct recordings, and correct voice recognition errors. 9. The speech recognition system according to claim 8, wherein feedback adjustment is performed on the speech recognition system.   The user moves a mouse pointer on the display device onto a waveform unit for feedback adjustment or points with a touch pad or a mouse to display the feedback selection item set on the display device. Item 10. The speech recognition system according to Item 9.   The feedback adjustment selection item included in the feedback adjustment selection item set includes “play”, “re-record”, “acceptance adjustment”, “change to handwritten input”, “change to keyboard input”, or any combination of the above. The speech recognition system according to claim 9.   Whether the system has a display device or can be connected to other display devices, a desktop personal computer, a notebook personal computer, a home multimedia system, a television, a DVD, an AV system, a portable or a PDA having a display device on a remote control The speech recognition system according to claim 1, wherein: The speech recognition system according to claim 9, wherein the word unit is a word, a classifier, or an idiom.

Images