JP2006023773A - Voice processing system - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a voice processing system which is simple and low-power consuming, and recognizes voices of a wearer without disturbing wearer's action. <P>SOLUTION: The voice processing system includes a head set with a wireless communication function, and an external device which can wireless communicate with the head set, and the head set with the wireless communication function is provided with a microphone for detecting the voice of the head set wearer and generating a voice signal, a voice recognition means for recognizing the voice signal and generating a discrimination signal corresponding to the content of the recognized voice signal, and a recognition result transmission means for sending the recognition signal generated by the voice recognition means to the external device by the wireless communication, and the external device performs an action corresponding to the received discrimination signal. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

  The present invention relates to a voice processing system, and in particular, a headset with a wireless communication function capable of simplifying operation of these functions and reducing power consumption while mounting a voice recognition function and a voice transmission function, and such a headset. The present invention relates to a voice processing technology required between a computer and a device equipped with a voice recognition function.

  Conventionally, in order to operate a device, it has been naturally necessary to operate a switch, a keyboard, and the like. As the operation of the device becomes more complicated, the number of switches increases and the operation sequence becomes complicated. In addition, there is an inconvenience that the switch and keyboard cannot be operated when both hands are occupied.

  In recent years, speech recognition technology has begun to be used as an effective means for solving these problems.

  The device using the voice recognition technology can control the operation of the device in response to the content of the voice uttered by the user of the device, so that the operation of the device can be greatly simplified. In addition, it is possible to control home appliances, machines, robots, etc. at remote locations by voice anytime and anywhere, and the mechanical (physical) switches can be reduced, so the economic effect is great and ubiquitous. Has attracted attention as an important technology of the times.

  Generally, in a device equipped with a voice recognition function for recognizing input voice, a user's voice is collected using a microphone provided in the device or a microphone connected by a cable. The device holds a reading of the vocabulary (recognition vocabulary) that is to be recognized by the device, and based on the reading, a word speech model that composes the corresponding recognition vocabulary is created in advance to recognize the input speech. Remember it. Recognition of input speech by this type of speech recognition apparatus is performed as follows.

  First, the sound signal detected by the microphone is acoustically analyzed to obtain a feature parameter series. Next, the input speech is recognized by comparing the feature parameter series of the obtained speech signal with a word speech model that forms each recognized vocabulary.

  In the speech recognition apparatus, when a microphone is installed in the device itself, if the user utters while leaving the device, noise is superimposed on the speech signal detected by the microphone, and the recognition performance deteriorates. Therefore, in order to recognize with high accuracy, the user must approach the device and speak. Even when the microphone is connected to the device via a cable, if the microphone is installed at a location away from the user, the microphone must be approached to speak after all.

  Although there is a close-talking microphone in which a microphone connected to the device is arranged near the user's mouth, there is a problem that a cable connecting the device and the microphone narrows the user's action range. When a wireless close-talking microphone is used, the user's behavior is not limited, but electrical noise is superimposed on the voice signal detected by the microphone, and voice recognition performance is degraded.

  Usually, in the speech recognition technique, a recognition result is output after performing a large amount of signal processing and collation processing. If these processes are not performed almost in real time, the device cannot perform a corresponding operation promptly after the user's utterance is completed. For this reason, it is necessary for a device equipped with a speech recognition technology to have sufficient calculation capability, and there is a problem that it is difficult to install in a device that is inexpensive or needs to be downsized.

  In recent years, portable electronic recording devices have begun to be used. This is for storing an audio signal built in the apparatus in a storage area in the apparatus and reproducing the stored audio, and is used for recording audio instead of a memo. Further, the stored voice can be transferred to a device such as a personal computer via a cable, and the voice data can be stored in a large-capacity hard disk mounted on the personal computer.

When the personal computer has a voice recognition function, the stored voice data can be recognized by voice recognition technology and converted into a text file.

  In a voice memo, voice recognition of a spoken sentence is performed by the normal voice recognition technique described above. That is, words that may be used in a sentence are selected in advance, and these words are used as a recognition vocabulary. In many cases, tens of thousands to 100,000 words are selected as such words, but if the topic is limited, the number may be smaller. A speech model of the corresponding word is created in advance from the reading of the recognized vocabulary and stored for the recognition of the input speech. Furthermore, a language model representing the ease of connection between these words is created in advance and stored for recognition of the input speech.

  In speech recognition, the accumulated speech data is acoustically analyzed to obtain a feature parameter series. Next, the input speech is recognized by comparing the obtained speech feature parameter series with the word speech model and language model of each recognition word created in advance.

  However, in the portable electronic recording device, in order to improve portability, the internal storage area is often configured by a semiconductor memory, and the amount of audio that can be stored inside is limited. In addition, when the stored voice is transferred to a personal computer or the like, it is necessary to connect with a cable or via a removable recording medium, and voice information cannot be transferred to other devices in real time.

  Further, when the device is used in a state where the hand is blocked, it is necessary to connect a headset type microphone or a microphone with a clip to the portable electronic recording device with a cable. In addition to obstructing behavior, cables are troublesome to connect each time.

  As described above, in a device using the conventional voice recognition technology, it is necessary to always pay attention to the positional relationship between the user and the microphone in order to recognize the voice accurately, and to speak near the microphone as necessary. there were.

  In addition, when using a headset type microphone, there is a problem that the action is hindered by a cable connecting the microphone and the device. With a headset that does not have the computational capacity required by voice recognition technology, voice operation itself is not possible.

  Moreover, in the portable electronic recording device, the amount of audio data that can be stored inside is limited, and the stored data cannot be transferred to other devices in real time. In addition, it is necessary to connect the microphone with a cable, which causes problems such as the cable obstructing the action and the connection is troublesome.

  The present invention provides a speech processing system capable of realizing a highly accurate speech recognition technique without interfering with user behavior in order to overcome the above-described problems.

  Also provided is a voice processing system including a headset with a wireless communication function that can transfer voice data to other devices in real time.

  Furthermore, a voice processing system including a headset with a wireless communication function capable of reducing power consumption by providing means for stopping the voice recognition function and voice transmission function when unnecessary by the function selection means is provided.

  Furthermore, the present invention provides an audio processing system capable of transferring audio data from a headset to a second device in real time and recognizing the audio by the second device. Furthermore, a speech processing system for controlling the operation of the third device by wirelessly transmitting the speech recognition result from the second device to the third device is provided.

In order to achieve the above object, according to the first aspect of the present invention, a headset with a wireless function includes:
(A) Microphone for detecting voice and generating a voice signal (b) Voice recognition means for recognizing the generated voice signal (c) Recognition result for sending the recognition result by the voice recognition means to an external device by wireless communication Transmission means (d) comprises function selection means for switching whether or not the generated voice signal is processed by the voice recognition means.

  Since there is no need to connect the headset to another device with a cable or the like, the user's action is not limited. Further, the user can arbitrarily select the voice recognition process by the function selection means. When the voice recognition process is selected, the recognition process is performed easily and with low power consumption in the headset with the wireless communication function. Even if voice recognition technology is not installed in an external device that can communicate wirelessly with the headset, these devices can be operated by voice commands, for example. In addition, simple speaker recognition, sentence recognition, dialogue understanding, and the like can be performed inside the headset.

In the second aspect of the present invention, a headset with a wireless communication function is:
(A) Microphone for detecting voice and generating a voice signal (b) Voice recognition means for recognizing the generated voice signal (c) Recognition result transmission for sending the recognition result by the voice recognition means to an external device by wireless communication Means (d) Voice transmission means for transmitting the generated voice signal to an external device by wireless communication (e) Function selection means for selecting whether the voice signal is processed by the voice recognition means or the voice transmission means Prepare.

  Preferably, the function selection unit further includes at least one of a mode in which the voice signal is not processed by either the voice recognition unit or the voice transmission unit, and a mode in which both the voice recognition unit and the voice transmission unit are processed.

  The user can arbitrarily select voice recognition processing and voice transmission processing by operating the function selection means. When voice recognition is selected, the voice is simply recognized with a small amount of calculation in the headset, for example, operating the remote device by the recognized voice command, as in the headset described in the first aspect. Can be recognized as a sentence. On the other hand, when audio transmission is selected, detailed audio recognition can be performed in the transmission destination device after the audio signal detected by the microphone is wirelessly transmitted. In this case, more accurate sentence recognition, intent understanding, speaker recognition, and dialogue understanding can be performed. In addition, when the audio data transmission destination device has a large-capacity storage device, the audio data over a long period of time can be constantly stored and played back, increasing usefulness.

  According to a third aspect of the present invention, there is provided a voice processing system including a headset with a wireless communication function and an external device capable of wireless communication with the headset. The headset with a wireless communication function constituting this system includes a microphone that detects a voice of a headset wearer and generates an audio signal, an identification corresponding to the content of the recognized audio signal, recognizing the generated audio signal Voice recognition means for generating a signal, and recognition result transmission means for sending the identification signal generated by the voice recognition means to the external device by wireless communication. On the other hand, when the external device receives the identification signal from the headset, the external device starts an operation corresponding to the identification signal.

  The external device has, for example, a table that stores a plurality of identification signals and operations corresponding to the respective identification signals in association with each other, and starts a desired operation by searching this table.

  With this voice processing system, an external device capable of wireless communication with the headset need only store the correspondence table, and hardly requires a structural change. A user wearing the headset can operate the external device using a voice command.

  In a fourth aspect of the present invention, a speech processing system includes a headset with a wireless communication function and an external device having a speech recognition function and capable of wireless communication with the headset. The headset with a wireless communication function includes a microphone that detects a voice of a wearer of the headset and generates an audio signal, and an audio transmission unit that transmits the audio signal to an external device by wireless communication. On the other hand, the external device includes a voice receiving unit that receives a voice signal transmitted from the headset, and a voice recognition unit that recognizes the received voice signal.

  For example, the voice recognition unit of the external device generates an identification signal corresponding to the content of the received voice signal, and the external device performs an operation corresponding to the generated identification signal.

  Alternatively, the voice recognition means converts the generated identification signal into a character string and outputs it. In this case, the external device further includes a display unit, and displays a character string as a voice recognition result.

  In this system, an external device has a voice recognition function. When the external device has a sufficient capacity and computing capacity, it is possible to perform voice recognition with a higher degree of difficulty.

  Further, by providing the external device with a text conversion function and a display function, it is possible to recognize characters in speech almost in real time while receiving a reception signal from the headset and display the recognition result on the screen.

  In a fifth aspect of the present invention, a speech processing system includes a headset with a wireless communication function, a first external device having a speech recognition function and capable of wireless communication with the headset, and a wireless communication with the first external device. A second external device capable of communication. The headset with a wireless communication function includes a microphone that detects a voice of a wearer of the headset and generates an audio signal, and an audio transmission unit that transmits the audio signal to the first external device by wireless communication. The first external device includes: a voice receiving unit that receives a voice signal transmitted from the headset; a voice recognition unit that recognizes the received voice and identifies an identification signal corresponding to the content of the recognized voice signal; And a recognition result transmission means for transmitting the identified signal to the second external device by wireless communication. The second external device performs an operation corresponding to the word ID received from the first external device.

  According to this system, the user's voice collected by the headset is recognized by using the first external device having a high capacity and computing capacity, and the second external device is connected via the first external device. Control the operation. As a result, more complicated audio processing can be performed.

  According to the present invention, the headset with wireless communication function is provided with voice recognition means, voice transmission means, and function selection means for switching between them, so that the voice according to the user's intention can be obtained without disturbing the user's action. A headset capable of recognition is provided.

  In the headset, simple and low power consumption voice recognition is performed, and when voice data is transmitted to a device outside the headset, more accurate voice recognition with high difficulty can be performed.

  Further, the voice recognition processing function and the voice transmission processing function can be arbitrarily paused by the user's selection, and the power consumption of the headset with the wireless communication function can be reduced.

  Furthermore, when the voice data is transferred from the headset to the second device having a large capacity, the second device recognizes the received voice in real time and enables text conversion, editing, storage, reproduction, and the like. This further improves the convenience of the system.

  In the present invention, a wireless headset equipped with a voice recognition function is positioned as the most human-friendly device in the wearable and ubiquitous era, expanding the performance and application of voice recognition, and reducing the size and cost of the headset. Is possible.

  In addition, the use of headsets and voice input that are most familiar to humans will accelerate the use of information equipment systems and networks for the elderly and persons with disabilities, as well as interaction with various equipment systems and various services and content. Can be used. As a result, it can contribute to the activation of various equipment system industries, information and communication media industries, and service industries.

  Embodiments of the present invention will be described below with reference to the drawings.

(First embodiment)
1 and 2 show an appearance of a headset 10 with a wireless communication function according to a first embodiment of the present invention and a schematic system configuration thereof. The headset 10 with a wireless communication function includes a microphone 13 that detects an audio generated by a wearer (user) of the headset 10 and generates an electrical audio signal, and an audio recognition that recognizes the audio signal through digital conversion. Unit 23, recognition result transmission means 25 for transmitting the recognition result by the voice recognition unit 23 from the wireless communication module 17 to an external device, and function selection for selecting whether or not the voice signal detected by the microphone 13 is subjected to voice recognition processing. Means 20 are provided. The function selection means includes a function selection switch 14, and the user can arbitrarily select a voice recognition process by operating the function selection switch 14.

  A headset with wireless communication function (hereinafter, simply referred to as “headset” in some cases) 10 has a shape in which left and right earpieces 11 are connected by a flexible frame, and is attached to a user's head. use. An arm 15 extends from one of the ears, and a microphone 13 is attached to the tip of the arm 15. When the user wears the headset 10, the microphone is located almost at the mouth of the user and detects voice with little superposition of ambient noise.

  In the ear pad 11, speakers (left and right) 17, a CPU board 16, a wireless communication module 17, and a battery 12 are incorporated. The function selection switch 14 is arranged outside either one of the ear contacts, and as described above, it is possible to select whether or not to perform the voice recognition processing with the intention of the user. Although not shown, each element is connected by a cable as necessary.

  The CPU board 16 includes a CPU and its peripheral circuits, a memory (not shown), an A / D converter 21, a function selection unit 19, and the like. The A / D converter 21 converts the analog audio signal detected by the microphone 13 into a digital audio signal, and inputs the conversion result to the CPU. The function selection unit 19 detects the state of the function selection switch 14 and notifies the CPU.

  The wireless communication module 17 performs digital wireless communication with an external device. More specifically, a transmission / reception function for transmitting a signal sent from the CPU board 16 to another external device (not shown), receiving a signal transmitted from the other device, and transferring the signal to the CPU board 16 have.

  The voice recognition means includes an A / D converter 21 and a voice recognition unit 23 on the CPU board 16. The audio transmission means 25 is realized by the CPU on the CPU board 16 and its peripheral circuits and the wireless communication module 17. The function selection means 20 is realized by the function selection switch 14, the CPU and peripheral circuits on the CPU board 16, and its output is connected to the voice recognition unit 23. As described above, when the user operates the function selection switch 14, the processing operation of the voice recognition unit can be controlled.

  The overview and system configuration of the headset 10 shown in FIGS. 1 and 2 are merely examples for realizing the technical idea of the present invention, and are not limited to such a configuration. For example, a circuit for performing dedicated voice recognition processing may be provided as voice recognition means. Further, for example, a DSP for performing signal processing at high speed may be provided. Further, for example, the function selection switch 14 may be divided into two and arranged at both ears.

  FIG. 3 shows an example of the function selection switch 14. The user can switch between the two states by operating the function selection switch 14 as necessary. Here, when the user selects to process the voice signal detected by the microphone 13 with the voice recognition unit 23, the state 1 is set, and when the user selects not to process, the state 2 is set.

  The function selection switch 14 is, for example, a switch of a type that has two push button switches and only one of them is always ON. When the user pushes the push button switch 31 to turn it on, the function selection switch 14 is in the state 1. In conjunction with this, the push button switch 32 is automatically turned OFF. Conversely, when the user presses the push button switch 32 to turn it on, the function selection switch 14 is in the state 2 and the other push button switch 31 is automatically turned off. The function selection unit 20 outputs a speech recognition operation signal to the speech recognition unit 23 in the state 1 according to the state of the function selection switch 14, and outputs a speech recognition stop signal to the speech recognition unit 23 in the state 2. .

  When the output of the function selection unit 19 is a voice recognition operation signal, the voice recognition unit 23 recognizes the voice signal detected by the microphone and sends the output to the recognition result transmission unit 25. When the output of the function selection unit 19 is a voice recognition stop signal, the operation is stopped.

  FIG. 4 shows the internal configuration of the voice recognition unit 23. The output of the A / D converter 21 is first input to the recognition signal blocker 41. The operation of the recognition signal blocker 41 is controlled by the output signal of the function selection unit 19. When the output of the function selection unit 19 is a voice recognition operation signal, the signal output from the A / D converter 21 is input to the acoustic analysis unit. When the output signal of the function selection unit is a voice recognition stop signal, the output from the A / D converter 21 is cut off.

  More specifically, when the output of the function selection unit 19 is a speech recognition operation signal, the recognition signal blocker 41 is closed, and the digital speech signal output from the A / D converter 21 is the acoustic analysis unit 43. Is input. The acoustic analysis unit 43 converts the input voice into feature parameters. As typical characteristic parameters used for speech recognition, a power spectrum that can be obtained by a band pass filter or Fourier transform, a cepstrum coefficient obtained by LPC (linear prediction) analysis, etc. are often used. The type of feature parameter does not matter. The acoustic analysis unit 43 converts the input speech into feature parameters at regular time intervals. Therefore, the output is a time series of characteristic parameters (characteristic parameter series). The feature parameter series is supplied to the model matching unit 45.

  On the other hand, in the recognized vocabulary storage unit 47, reading information of words necessary for creating a speech model of each word constituting the recognized vocabulary and an identifier corresponding to the recognition result when each word is recognized, for example, a command ID is stored. In the present embodiment, voice control by word recognition is described as an example of voice recognition in the headset, but the present invention is not limited to this. The voice recognition unit 23 in the headset performs voice recognition with a small amount of computation, memory capacity and power consumption, such as continuous word recognition, sentence recognition, word spotting, and voice intention understanding, and the result is transmitted to an external device system by wireless communication. Can be transmitted.

  The recognition model creation / storage unit 49 outputs the speech model of each word according to the recognition vocabulary stored in the recognition vocabulary storage unit 47 and the collation unit 45 as a recognition result when each word becomes a recognition result. A word ID as an identification signal is stored in advance. Of course, when recognition other than word recognition is performed, an identification signal corresponding to the recognition is stored.

  The model matching unit 45 obtains the similarity or distance between each speech model of the word to be recognized stored in the speech model creation / storage unit 49 and the feature parameter series of the input speech, and the similarity is maximum. The word ID associated with the speech model (or the smallest distance) is output as a recognition result.

  As a matching method of the model matching unit 45, a voice model is also expressed by a feature parameter series, and a distance between the feature parameter series of the voice model and the feature parameter series of the input voice is obtained by DP (dynamic programming), A method of expressing a speech model using an HMM (Hidden Markov Model) and calculating the probability of each speech model when a feature parameter sequence of the input speech is input is widely used. It doesn't matter.

  The word ID output from the model matching unit 45 is directly output from the speech recognition unit 23 and input to the recognition result transmission unit 25 (see FIG. 2). The recognition result transmission unit 25 wirelessly transmits the word ID to another device using the transmission function of the wireless communication module 17.

  When the output of the function selection unit 19 is a voice recognition stop signal, the recognition signal blocker 41 is open, and the A / D signal is not input to the acoustic analysis unit 43. Therefore, there is no output from the acoustic analysis unit 43. Similarly, since there is no input to the model matching unit 45, there is no output from the model matching unit 45.

  As described above, when the user of the headset 10 selects not to perform processing by the voice recognition means (that is, when the function selection switch 14 is in the state 2), the acoustic analysis unit 43, the model matching unit 45, and the recognition result. A series of processing by the transmission means 25 is not performed. In this case, the calculation amount is greatly reduced. When the CPU realizing the acoustic analysis unit 43, the model matching unit 45, and the recognition result transmission unit 25 has a power saving mode for temporarily reducing the computing capacity and power consumption, the state of the function selection switch 14 When the state becomes the state 2 or when the voice recognition stop signal is detected, the CPU can be shifted to the power saving mode. While the user chooses not to process the voice signal with the voice recognition means, the CPU operates in the power saving mode, so the load on the battery is reduced and the operating time of the headset with wireless communication function is extended. Can do. When the function selection switch 14 leaves the state 2 (that is, when a voice recognition operation signal is output), the CPU is immediately shifted to the normal mode so that the original computing ability can be exhibited.

  FIG. 5 shows an example of the stored contents of the recognized vocabulary storage unit 47 provided in the headset. In this example, the user wearing the headset 10 controls the air conditioner with a voice command. Accordingly, the result of the voice recognition unit 23 recognizing the voice uttered by the user is transmitted to the air conditioner by wireless communication.

  In the example of FIG. 5, “Eakontsu”, “Eatontome”, “Ondo Rae”, and “Ondo Sagaru” are stored as recognition vocabulary, and “01” and “02” are stored in each vocabulary. , “03”, “04”. When the voice recognition unit 23 of the headset 10 recognizes the voice “turn on air conditioner” issued by the user, the ID “01” is wirelessly transmitted to the air conditioner.

  In accordance with the stored content of the recognized vocabulary storage unit 47, the stored content of the speech model creation / storage unit 49 is created. In the example of stored contents in Fig. 5, acoustic models corresponding to the words "Eakontsu", "Eatontome", "Ondo Rae", and "Ondo Saeru" are created. A pair with an identification signal (word ID) is stored.

  On the other hand, as shown in FIG. 6, the air conditioner stores each word ID in combination with the corresponding operation. Therefore, when a voice recognition result (ie, word ID) is received from the headset, an operation corresponding to the word ID is performed.

  FIG. 7A shows a case where the user of the headset utters “turn on the air conditioner” in a state where the voice recognition processing mode is selected by the function changeover switch 14. The voice uttered by the user is detected by a microphone and converted into a digital signal by the A / D converter 21. Since the state of the function selection switch 14 is state 1, the function selection unit 19 outputs a voice recognition operation signal. Accordingly, the recognition signal blocker 41 is closed, and the digital signal is input to the voice analysis unit 43, converted into a feature parameter series, and input to the collation unit 45. The collation unit 45 collates the input feature parameter series with the speech model of each word stored in the acoustic model creation / storage unit 49. As a result, when the similarity of the speech model corresponding to “Eakontsu” is the highest, the collation unit 45 outputs the word ID “01” as the recognition result.

  The word ID “01” is input to the recognition result transmission means 25, and the word ID “01” is transmitted to the air conditioner.

  When the air conditioner receives the word ID “01”, the operation of the air conditioner function is started according to the correspondence table of FIG. FIG. 7B shows a case where the user of the headset utters “turn on air conditioner” in a state where the function changeover switch 14 selects a mode in which voice recognition processing is not performed. The voice uttered by the user is detected by a microphone and converted into a digital signal by the A / D converter 21. Since the function selection switch 14 is in the state 2, the function selection means 19 outputs a voice recognition stop signal. Therefore, the recognition signal blocker 41 is open, and the digital signal is not input to the voice analysis unit 43. In this case, the recognition result is not obtained and the recognition result is not transmitted to the air conditioner. The air conditioner does not start operation.

  The above-described headset 10 with the wireless communication function detects the user's voice using the attached microphone 13. Since the attached microphone 13 is disposed in the vicinity of the user's mouth, the sound signal detected by the microphone has little superposition of ambient noise, and high recognition performance can be obtained when recognizing the sound.

  Since the recognized voice command is transmitted to another device by wireless communication, a cable is not required and the user's action is not hindered.

  Since voice recognition is performed on the headset 10 side, a device having a function capable of wireless communication with the headset can be operated with voice generated by the user without installing voice recognition technology.

  Further, since the function selection means for selecting whether or not to process by the voice recognition means is provided, the user can select not to perform the voice recognition process on the voice he / she made by his / her own intention. While the voice recognition means is operating, it is necessary to drive the arithmetic unit with a high-speed operation clock in order to process a large amount of calculation in real time and process the detected voice signal, but when the voice recognition means does not process the voice Can eliminate the need for calculations related to speech recognition, and can reduce the operation clock of the arithmetic unit. Since the arithmetic device has higher power consumption as the operation clock is higher, it is possible to significantly reduce the power consumption of the headset with the wireless communication function by stopping the processing in the voice recognition means. The headset with a wireless communication function cannot be supplied with power from the outside, and operates with a battery or a storage battery. Therefore, when the power consumption is reduced, the operation time of the headset with the wireless communication function can be extended, and the usefulness of the headset with the wireless communication function is improved.

(Second embodiment)
FIG. 8 shows a system configuration example of a headset according to the second embodiment of the present invention. In the first embodiment, a voice signal is simply analyzed and verified by a voice recognition unit, and an identification (ID) signal corresponding to a vocabulary issued by a user is wirelessly transmitted to an external device to be controlled. Indicated. In the second embodiment, a configuration example in which voice data before voice recognition is wirelessly transmitted to other devices in real time in addition to voice recognition in the headset will be described.

  First, an audio signal detected by the microphone 13 is input to the A / D converter 21 and converted from an analog signal to a digital audio signal. The digital voice signal is divided into two parts, one is input to the voice recognition unit 23 and the other is input to the voice transmission means 53.

  The function selection unit 50 includes a function selection switch 51 and a function selection unit 19. By operating the function selection switch 51, the user can switch between two states as necessary. Here, state 1 is selected when the speech signal detected by the microphone is processed by the speech recognition unit 23, and state 2 is selected when the speech signal detected by the microphone is processed by the speech transmission means 53. Will be.

  FIG. 9 shows an example of the function selection switch 51. The function selection switch 51 has two push button switches. Only one of these two push button switches is always ON. When the user presses the push button switch 51 to turn it on, the function selection switch is in the state 1. In conjunction with this, the push button switch 101 is automatically turned OFF. When the user presses the push button switch 102 to turn it ON, the function selection switch is in the state 2. In conjunction with this, the push button switch 101 is automatically turned OFF. When the function selection switch 51 is in the state 1, the function selection unit 19 outputs a voice recognition operation signal to the voice recognition unit 23 and simultaneously outputs a voice transmission stop signal to the voice transmission unit 53. When the function selection switch 51 is in the state 2, a voice recognition stop signal is output to the voice recognition unit 23, and at the same time, a voice transmission operation signal is output to the voice transmission unit 53. The operation of the voice recognition unit 23 is the same as that described in the first embodiment.

  FIG. 10 shows an internal configuration of the voice transmission unit 53.

  The audio signal converted into a digital signal by the A / D converter 21 is first input to the transmission signal blocker 55. The transmission signal blocker 55 is closed when the output signal from the function selection unit 19 is a transmission operation signal, and the signal output from the A / D converter 21 is input to the speech encoding unit 57. When the output signal of the function selection unit 19 is a transmission stop signal, the transmission signal breaker 55 is opened and the output from the A / D converter 21 is blocked.

  The voice encoding unit 57 encodes the digital voice signal input via the transmission circuit breaker 55 by a predetermined method. As processing for encoding a digital audio signal, compression processing by ADPCM or the like, information addition for correcting encoding parameters and transmission errors, and the like can be considered, but the specific processing contents are not limited here.

  The encoded data is input to the audio transmission unit 59. The audio transmission unit 59 wirelessly transmits the encoded data to another device using the transmission function of the wireless module 17 (FIG. 1).

  FIG. 11 shows a specific operation of the headset with a wireless communication function according to the second embodiment. Here, an example will be described in which a user wirelessly controls both an air conditioner and a personal computer in a room using a headset with a wireless communication function. The user's voice collected by the microphone is wirelessly transmitted to the air conditioner as an output of the headset recognition result transmission means 25, and on the other hand, to the personal computer as an output (encoded data) of the voice transmission means 53. Sent.

  The storage contents of the recognition vocabulary storage unit 47 and the speech model creation / storage unit 49 of the speech recognition unit 23 in the headset and the setting storage content on the air conditioner side are the same as those in the first embodiment. Further, it is assumed that a large-capacity hard disk is connected to the personal computer, and all audio data received from the headset with the wireless communication function is stored in this hard disk.

  In the example of FIG. 11A, the user has uttered a voice command “Electrify” with the function changeover switch 51 set to the voice recognition mode. The voice uttered by the user is detected by a microphone and converted into a digital signal by the A / D converter 21. The digital signal is divided into two, and as described above, one is input to the voice recognition unit 23 and the other is input to the voice transmission means 53.

  At this time, since the function selection switch 51 is in state 1, the function selection unit 19 outputs a voice recognition operation signal to the voice recognition unit 23 and outputs a voice transmission stop signal to the voice transmission means 53.

  The digital signal input to the voice recognition unit 23 is first input to the recognition signal blocker 41. Since the recognition signal blocker 41 is closed by the voice recognition operation signal from the function selection unit 19, the digital signal is input to the acoustic analysis unit 43 as it is. Processing after the verification is the same as that in the first embodiment. That is, the identification signal “01” is output as the recognition result from the model matching unit 45, and the signal “01” is wirelessly transmitted from the recognition result transmission unit 25 to the air conditioner.

  On the other hand, the digital signal input to the voice transmission means 53 is input to the transmission signal blocker 55. Since the function selection unit 19 outputs a voice transmission stop signal, the transmission signal breaker is open. Therefore, the digital signal is not input to the speech encoding unit, and the subsequent processing is not performed.

  FIG. 11 (b) shows that the user uttered “I will talk about music today” in the state where the voice transmission means processing mode is selected with the function changeover switch 51. The voice uttered by the user is detected by a microphone and converted into a digital signal by the A / D converter 21. The digital signal is divided into two parts, one is input to the voice recognition unit 23 and the other is input to the voice transmission means 53.

  Since the function selection switch 51 is in the state 2, the function selection unit 19 outputs a voice recognition stop signal to the voice recognition unit 23 and outputs a voice transmission operation signal to the voice transmission unit 53.

  The digital signal input to the voice recognition unit 23 is first input to the recognition signal blocker 41. However, since the function selection unit 19 outputs a voice recognition stop signal, the recognition signal blocker 41 is open. is there. Therefore, the digital signal is not input to the acoustic analysis unit 43, and the subsequent processing is not performed.

  On the other hand, the digital signal input to the audio transmission means 53 is first input to the transmission signal blocker 55. Since the function selection unit outputs the audio transmission operation signal, the transmission signal blocker 55 is closed. Therefore, the digital signal is encoded by the audio encoding unit 57 and wirelessly transmitted from the audio transmission unit 59 to the personal computer via the wireless communication module 17.

  The personal computer decodes the encoded audio sent from the headset, returns it to a digital audio signal, and records it on the hard disk. That is, the content that the user has spoken is recorded in the personal computer by wireless communication from the headset. Since the capacity of the personal computer is sufficient, the content spoken by the user can be stored as speech or in a text-converted state. The recorded voice can be searched and reproduced as appropriate.

  Further, as will be described later, when a voice recognition function is provided in a personal computer, it is possible to perform an accurate voice recognition process with a high degree of difficulty using a voice signal transmitted from a headset.

  With such a configuration, a user wearing a headset with a wireless function can perform audio processing on a plurality of devices in a hands-free state according to his / her selection. For example, it is possible not only to control other devices by voice commands, but also to record the content of what you have spoken in real time.

(Third embodiment)
12 and 13 schematically show the system configuration of a headset with a wireless function according to the third embodiment of the present invention.

  In the third embodiment, as in the second embodiment, the voice signal can be processed by both voice recognition processing for voice commands and transmission processing for wireless transmission of voice data. In the third embodiment, in addition to these two processing modes, an OFF mode that does not process either is added to the function selection switch.

  As shown in FIGS. 12 and 13, the function selection means 60 includes a function selection switch 61 and a function selection unit 19. The user can switch between the three states with the function selection switch 61 as necessary. State 1 when the user has selected speech recognition processing of the voice he / she uttered, state 2 when the user has selected speech transmission processing, and speech is not processed by the speech recognition means or the speech transmission means. If selected, state 3 is assumed.

  An example of the function selection switch 61 is shown in FIG. The function selection switch 61 is provided with three push button switches, and only one of these three buttons is always in an ON state. When the user presses the push button switch 101 to turn on the voice recognition, the function selection switch 61 is in the state 1. In conjunction with this, the push button switches 102 and 103 are automatically turned off. When the user presses the push button switch 102 to turn on the voice transmission, the function selection switch 61 is in the state 2, and the push button switches 101 and 103 are automatically turned off in conjunction with this. When the push button switch 103 is pressed, the function selection switch 61 is in the state 3, and in conjunction with this, the push button switches 101 and 102 are automatically turned off.

  When the state of the function selection switch 61 is state 1, the function selection unit 19 outputs a voice recognition operation signal to the voice recognition unit 23 and simultaneously outputs a voice transmission stop signal to the voice transmission unit 53. When the state of the function selection switch 61 is state 2, a voice recognition stop signal is output to the voice recognition unit 23, and at the same time, a voice transmission operation signal is output to the voice transmission unit 53. When the state of the function selection switch 61 is state 3, a voice recognition stop signal is output to the voice recognition unit 23, and at the same time, a voice transmission stop signal is output to the voice transmission means 53.

  The operation of the voice recognition unit 23 is the same as in the first and second embodiments, and the operation of the voice transmission unit 53 is the same as in the second embodiment.

  When the user selects not to perform processing in either the voice recognition unit 23 or the voice transmission means 53, that is, when the function selection switch 61 is in the state 3, the recognition breaker 41, the voice recognition stop signal, and the voice transmission stop signal, Both transmission circuit breakers 55 are open. Therefore, the processing of the acoustic analysis unit 43, the model matching unit 45, the recognition result transmission unit 25, the speech encoding unit 57, and the speech transmission unit 59 is not performed, and the amount of calculation is greatly reduced.

  When the CPU that implements the acoustic analysis unit 43, the model matching unit 45, the voice encoding unit 57, and the voice transmission unit 59 has the power saving mode, the user selects the OFF mode (that is, the function selection switch 61 is When the state 3 is entered, or when a voice recognition stop signal and a voice transmission stop signal are detected, the CPU can be shifted to the power saving mode. In the power saving mode, it is possible to save power by reducing the computing capacity and power consumption of the CPU. Therefore, the load on the battery is reduced, and the operation time of the headset can be extended. When the function selection switch 61 is removed from the state 3 or when at least one of the voice recognition operation signal and the voice transmission operation signal is output, the CPU can be promptly shifted to the normal mode in which the original computing ability can be exhibited. Good.

  14 and 15 illustrate specific operations of the headset with the wireless communication function according to the third embodiment. As in the second embodiment, it is assumed that a user wearing a headset performs control by voice commands or transmission of voice data to an indoor air conditioner and a personal computer.

  The storage contents of the recognition vocabulary storage unit 47 and the speech model creation / storage unit 49 of the speech recognition unit 23 and the table setting of the air conditioner are the same as in the first and second embodiments. Similarly to the second embodiment, a large-capacity hard disk is connected to the personal computer, and all audio data received from the headset with the wireless communication function is stored in this hard disk.

  FIG. 14A shows a state where the user selects the voice recognition mode with the function selection switch 61 and utters a voice command “Easy-on” toward the microphone. The user's voice is detected by a microphone and converted into a digital signal by the A / D converter 21. The digital signal is divided into two parts, one is input to the voice recognition unit 23 and the other is input to the voice transmission means 53. Since the function selection switch 61 is in the state 1, the function selection unit 19 outputs a voice recognition operation signal to the voice recognition unit 23 and outputs a voice transmission stop signal to the voice transmission unit 53. In this case, as in the second embodiment (FIG. 11A), the command “01” is wirelessly transmitted to the air conditioner, and the air conditioner starts operation. On the other hand, audio data is not transferred to the personal computer.

  FIG. 14B shows a state where the user utters “I will talk about music today” in a state where the audio transmission mode is selected with the function changeover switch 61. The voice uttered by the user is detected by a microphone and converted into a digital signal by the A / D converter 21. The digital signal is divided into two parts, one is input to the voice recognition unit 23 and the other is input to the voice transmission means 53.

  Since the function selection switch 61 is in the state 2, the function selection unit 19 outputs a voice recognition stop signal to the voice recognition unit 23 and outputs a voice transmission operation signal to the voice transmission unit 53. At this time, as in the second embodiment (FIG. 11B), nothing is transmitted to the air conditioner, but an encoded audio signal is transmitted to the personal computer. As a result, the user can record the contents spoken by the user, for example, in a memory in the PC. If a command vocabulary and word ID table is also set on the personal computer side, when recording, the user wirelessly transmits a voice command that has undergone voice recognition processing to the personal computer, and turns on the computer. Is also possible.

  FIG. 15 shows a state in which the user utters “Today I will talk about music” in a state where the function selector switch 61 is selected to be in the OFF mode, that is, neither voice recognition nor voice transmission processing is performed. The voice uttered by the user is detected by a microphone and converted into a digital signal by the A / D converter 21. The digital signal is divided into two parts, one is input to the voice recognition unit 23 and the other is input to the voice transmission means 53.

  Since the function selection switch 61 is in the state 3, the function selection unit 19 outputs a voice recognition stop signal to the voice recognition unit 23 and outputs a voice transmission stop signal to the voice transmission unit 53.

  The digital signal input to the voice recognition means 23 is first input to the recognition signal blocker 41. However, since the function selection unit 19 outputs a voice recognition stop signal, the recognition signal blocker 41 is open. is there. Therefore, the digital signal is not input to the acoustic analysis unit 43, and the subsequent processing is not performed.

  Similarly, the digital signal input to the voice transmission means 53 is first input to the transmission signal blocker 55. However, since the function selection unit 19 outputs the voice transmission stop signal, the transmission signal blocker 55 is output. Is also open. Therefore, the digital signal is not input to the audio encoding unit 57, and the subsequent processing is not performed.

  Therefore, no audio control signal is sent to the air conditioner, and no audio data is sent to the personal computer. However, the user can use functions that are not intended for voice recognition processing and operations associated therewith, for example, control or dictation of other devices. Therefore, the user can listen to music or a third party's voice through the speaker built in the headset.

(Fourth embodiment)
16 and 17 schematically show the system configuration of a headset with a wireless communication function according to the fourth embodiment of the present invention.

  The sound detected by the microphone 13 is input to the A / D converter 21 and converted from an analog signal to a digital sound signal. The digital voice signal is divided into two parts, one is input to the voice recognition unit 23 and the other is input to the voice transmission means 53.

  The function selection unit 70 includes a function selection switch 71 and a function selection unit 19. The function selection switch 71 can switch between three states by a user operation. State 1 when the user has selected to process the audio signal detected by the microphone 13 with the speech recognition unit 23, and state when the user has selected to process the audio signal detected by the microphone 13 with the audio transmission means 53. 2. If it is selected that the voice signal detected by the microphone 13 is to be processed by both the voice recognition unit 23 and the voice transmission means 53, state 3 is set.

  FIG. 17 shows an example of the function selection switch 71. The function selection switch 71 has three push button switches: a voice recognition button 101, a voice transmission button 102, and both mode buttons 104. Only one of these push button switches is always ON. When the user turns on the push button switch 101, the function selection switch 71 is in the state 1, and in conjunction with this, the push button switches 102 and 104 are automatically turned off. Similarly, when the user turns on the push button switch 102, the function selection switch 71 is in the state 2, and the push button switches 101 and 104 are automatically turned off in conjunction with this. When the push button switch 104 is turned on, the function selection switch 71 is in the state 3, and in conjunction with this, the push button switches 101 and 102 are automatically turned off.

  When the function selection switch 71 is in the state 1, the function selection unit 19 outputs a voice recognition operation signal to the voice recognition unit 23 and outputs a voice transmission stop signal to the voice transmission unit 53. When the function selection switch 71 is in the state 2, a voice recognition stop signal is output to the voice recognition unit 23, and a voice transmission operation signal is output to the voice transmission means 53. When the function selection switch 71 is in the state 3, a voice recognition operation signal is output to the voice recognition unit 23, and at the same time, a voice transmission operation signal is output to the voice transmission means 53.

  The operations of the voice recognition unit 23 and the voice transmission unit 53 are the same as those in the above-described embodiment.

  FIG. 18 is a diagram for explaining a specific operation of the headset with the wireless communication function of FIG. In the example shown in FIGS. 18A and 18B, as in the third embodiment, the user wearing the headset with the wireless communication function switches and selects the voice recognition mode and the voice transmission mode with the function selection switch 71. The voice control of the air conditioner and the transmission and recording of the voice data to the personal computer are performed. The stored contents of the recognized vocabulary storage unit of the headset and the speech model creation / storage unit are the same as in the example of the first embodiment. The settings on the air conditioner side are the same as in the first embodiment. Also, a large-capacity hard disk is connected to the personal computer, and all audio data received from the headset with wireless communication function is stored in this hard disk. Shall be.

  FIG. 19 shows a state in which the user has selected to process the voice in both voice recognition and voice transmission with the function selector switch 71. The user wearing the headset has just voiced “Turn on air conditioning”. The voice uttered by the user is detected by the microphone 13 and converted into a digital signal by the A / D converter 21. The digital signal is divided into two parts, one is input to the voice recognition unit 23 and the other is input to the voice transmission means 53.

  Since the function selection switch 71 is in the state 3, the function selection unit 19 outputs a speech recognition operation signal to the speech recognition unit 23 and outputs a transmission operation signal to the speech transmission means 53.

  The digital signal input to the voice recognition unit 23 is first input to the recognition signal blocker 41. Since the function selection unit 19 outputs a voice recognition operation signal, the recognition signal blocker 41 is closed. The digital audio signal is input to the acoustic analysis unit, the recognition result “01” is wirelessly transmitted to the air conditioner, and the air conditioner starts operation.

  On the other hand, the digital signal input to the audio transmission means 53 is first input to the transmission signal blocker 55. Since the function selection unit 19 outputs the audio transmission operation signal, the transmission signal breaker 55 is also closed. The digital audio signal is input to the audio encoding unit, and the encoded audio signal is wirelessly transmitted to the personal computer.

  In this case, the voice data stored in the personal computer includes a voice component uttered in expectation of being recognized by the voice recognition unit 23 of the headset with the wireless communication function. Therefore, it is possible to check the operation history of the voice recognition unit 23 by playing back the voice stored in the personal computer.

  In the fourth embodiment, a voice uttered by a user is recognized as a voice command for device control, and at the same time, processed as voice data recorded and stored in a personal computer. A headset having such a configuration, for example, in a laboratory or factory can remotely control devices and equipment with voice commands without key operations, and simultaneously record the operation control records on a computer or the like. . The voice recognition processing in the headset is exemplified by voice command processing based on word recognition. However, as described above, the voice recognition of the headset of the present invention is not limited to this.

(Fifth embodiment)
FIG. 20 shows an outline of the system configuration of a headset with a wireless communication function according to the fifth embodiment of the present invention. The fifth embodiment is a combination of the third embodiment and the fourth embodiment described above, and the function selection switch has four modes: a voice recognition mode, a voice transmission mode, an OFF mode, and a voice recognition / transmission mode. Have.

  As in the third and fourth embodiments, the sound detected by the microphone 13 is input to the A / D converter 21 and converted from an analog signal to a digital sound signal. The digital voice signal is divided into two parts, one is input to the voice recognition unit 23 and the other is input to the voice transmission means 53.

  The function selection unit 80 includes a function selection switch 81 and a function selection unit 19. The function selection switch 81 can switch between four states according to the user's selection. If the user chooses to process the voice signal detected by the microphone 13 with the voice recognition unit 23, the state 1, if the user chooses to process with the voice transmission unit 53, the voice recognition unit 23 and the voice transmission. State 3 is selected when processing is performed by both of the means 53, and state 4 is selected when processing is not performed in either case.

  FIG. 21 shows an example of the function selection switch 81. The function selection switch 81 has four push button switches, and only one of these four push button switches is always ON. When the user turns on the push button switch 101, the function selection switch 81 is in the state 1, and in conjunction with this, the other three push button switches 102, 103, 104 are automatically turned off. Similarly, when any one is selected, the other three are automatically turned OFF.

  The signal output state of the function selection unit 19 corresponding to the state (mode) of the function selection switch 81, the operation of the signal breakers 41 and 55, and the word ID transmitted wirelessly are the same as in the third and fourth embodiments. Since it is the same, description is abbreviate | omitted here.

  22 and 23 show examples of specific operations of the headset with the wireless communication function shown in FIG. The user wearing the headset can select the four modes as appropriate by operating the function selection switch 81. 22A and 22B show an example in which the voice recognition mode and the voice transmission mode are switched to switch the control of the air conditioner by voice commands and the transmission and storage of voice data to a personal computer. 23 (a) and 23 (b) show examples of a mode in which both voice recognition and voice transmission are performed simultaneously and a mode in which neither is performed. As described in the third and fourth embodiments, in the mode in which both are performed, the air conditioner is controlled by a voice command, and at the same time, the voice is wirelessly transmitted to the personal computer as encoded data and stored. The stored data can be reproduced and analyzed later. In the OFF mode, neither voice recognition nor voice transmission is performed, but the user can listen to music or a third party's voice through a speaker built in the headset.

  The recognized vocabulary storage unit in the headset, the storage contents of the voice model creation / storage unit, and the storage and setting of the air conditioner are the same as in the first embodiment. A large-capacity hard disk is connected to the personal computer, and all audio data received from the headset with the wireless communication function is stored in this hard disk.

(Sixth embodiment)
FIG. 24 shows an outline of a speech processing system according to the sixth embodiment of the present invention. This voice processing system includes the headset 110 with the wireless communication function described in the first to fifth embodiments and the device 130 with the voice recognition function. In this system, when the voice transmission mode is selected by the function selection switch 114 of the headset, the voice signal detected by the microphone is transmitted to the voice recognition function device 130 via the voice transmission means 153 of the headset. Wireless transmission is performed, and voice recognition processing is performed on the device side. When the voice recognition mode is selected in the headset, voice recognition processing is performed in the headset.

  That is, the headset 110 with the wireless communication function includes a microphone 113 that detects a user's voice, a voice recognition unit that performs a process of recognizing the voice detected by the microphone 113, and a recognition result transmission unit 125 that wirelessly transmits the recognition result. The voice transmission means 153 wirelessly transmits the voice signal detected by the microphone 113 as encoded voice data, and the function selection switch 114 that selects either voice recognition or voice transmission processing.

  On the other hand, the device 130 with the voice recognition function includes a voice receiving unit 140 that receives voice data wirelessly transmitted from the headset, and a voice recognition engine 150 that recognizes the received voice.

  FIG. 25 shows the voice receiving means 140 of the device 130 with the voice recognition function shown in FIG. The encoded audio signal transmitted from the headset through wireless communication is received by the encoded audio receiving unit 141 and input to the encoded audio decoding unit 143.

  The encoded speech decoding unit 143 performs a decoding process on the encoded speech and outputs a digital speech signal to the speech recognition engine 150.

  The speech recognition engine 150 may use either word speech recognition technology or large vocabulary sentence speech recognition technology. Here, the configuration when the large vocabulary sentence speech recognition technology is used will be described.

  FIG. 26 is a schematic diagram of a speech recognition engine 150 using sentence speech recognition technology. The speech recognition engine 150 collects vocabulary that may be used in the input speech beforehand. For example, when the vocabulary is in units of words, the notation, reading, and word ID of each word are stored in the recognized vocabulary storage unit 157. Usually, about tens of thousands to 100,000 words are stored as such words, but when the topic or sentence pattern can be limited, the number of words can be narrowed down to reduce the storage capacity.

  In addition, a language model representing ease of connection between the words stored in the recognized vocabulary storage unit 157 in advance is created and stored in the language model storage unit 161. As the language model, for example, probability values created based on the appearance frequency of words in a sentence database collected in large quantities, the appearance frequency of two word combinations, and three word combinations can be used.

  The speech model creation / storage unit 159 generates a word speech model from the reading of each word stored in the recognized vocabulary storage unit 157, and stores it as a pair with the word ID of the word. Here, the generally well-known HMM (Hidden Markov Model) is often used as the word speech model, but it is not limited to this.

  The acoustic analysis unit 151 converts the input voice into feature parameters. As typical feature parameters used for speech recognition, a power spectrum that can be obtained by a band-pass filter or Fourier transform, or a cepstrum coefficient obtained by LPC (linear prediction) analysis is often used. The type of feature parameter does not matter. The acoustic analysis unit converts the feature parameter of the input speech at regular time intervals. Therefore, the output is a time series of characteristic parameters (characteristic parameter series).

  The model matching unit 155 obtains the similarity or distance between the continuous word speech model connected to each speech model of the word stored in the speech model creation / storage unit 159 and the input feature parameter series, and the acoustic similarity Calculate (distance). Further, the sequence of each word constituting the continuous word speech model is collated with each language model stored in the language model storage unit 161, and the linguistic accuracy is calculated. The model matching unit 155 considers acoustic similarity and linguistic accuracy, obtains a word sequence that best matches the input feature parameter series, and a word ID sequence of words constituting the word sequence Is output to the word ID notation conversion unit 163 as a recognition result.

  The word ID notation conversion unit 163 collates the word ID series with the word IDs and notations stored in the recognized vocabulary storage unit 157 and concatenates the notations to convert them into character strings corresponding to the word ID series.

  FIG. 27 illustrates a specific operation of the voice processing system shown in FIGS. In the example of FIG. 27, a user wearing a headset with a wireless communication function selects the voice transmission mode with the function selection switch 114 and transfers the voice he / she speaks to a device with a voice recognition function (personal computer).

  The voice “speaking about music today” uttered by the user is detected by the microphone 113, encoded, and transferred from the voice transmission means 153 to the personal computer. The personal computer decodes the received signal and performs voice recognition processing. On the computer side, word notation, reading, and word ID are stored in advance in the recognition vocabulary storage unit 157 of the speech recognition engine 150.

  FIG. 28 shows an example of stored contents of the recognized vocabulary storage unit 157. For example, a reading “ongaku” and a word ID “00811” are registered corresponding to the notation “music”. The speech model creation / storage unit 159 creates and stores a word speech model corresponding to “music” or the like according to the stored contents of the recognized vocabulary storage unit 157.

  FIG. 29 shows an example of the contents stored in the language model storage unit 161. In the stored content example shown in FIG. 29, the first word ID, the second word ID that continues immediately after the first word ID, and the second word ID that immediately follows the word indicated by the first word ID are indicated. The degree of appearance of words (ease of appearance) is stored in association with each other. For example, the degree (ease of appearance) that the word ID 00712 and the word ID 00811 are used in succession is 0.012. The degree (ease of appearance) that the word ID 02155 is used after the word ID 00712 is 0.584.

  If the stored contents of the recognized vocabulary storage unit 157 are collated, it can be understood that the combinations of the above-described word IDs represent “to”, “music”, “to” and “to do”. Further, when referring to the ease of appearance, it can be seen that the latter combination has a higher probability of appearing continuously than the former. Therefore, the character string “I do” is preferentially selected.

  25 and 26, the voice transferred from the headset is first received by the coded voice receiving unit 141 of the personal computer, decoded into a voice signal by the coded voice decoding unit 143, and then the voice recognition engine 150. Is input.

  The decoded speech signal is converted into a feature parameter series by the acoustic analysis unit 151 and input to the model matching unit 155. Based on the speech model of each word stored in the speech model creation / storage unit 159 and the language model stored in the language model storage unit 161, the model matching unit 155 obtains a sequence of word IDs corresponding to the parameter sequence. . In this case, the obtained word ID series is “01211, 12322, 00811, 08211, 12596, 00712, 02155”.

  The word ID notation conversion unit 163 obtains a character string “Today I will talk about music” by obtaining a notation corresponding to each word ID of the word ID series and concatenating them.

  In the case where the device with speech recognition function 130 has a function of displaying characters, by displaying the character string converted by the model matching unit 155 on the device with speech recognition function 130, the user can use the contents spoken by the user as characters. You can check on the spot. FIG. 30 shows an example in which the personal computer displays the character string as text in this way.

  Further, when the device 130 with the voice recognition function has an editing function, real-time editing can be performed on the spot. In this case, the work efficiency is remarkably improved as compared with the case where the voice signal is accumulated and converted into a character string and edited later.

  Further, the function selection switch 114 of the headset 110 with the wireless communication function is switched so as to be recognized by the voice recognition unit 123 of the headset itself, and the command voice for editing is recognized there, and the recognition result is converted into a device with the voice recognition function. If wireless transmission is performed at 130, editing work can be performed by voice. Since the function selection switch 114 is provided in the headset, the trouble of switching the processing mode is not a problem here. It is possible to omit switching of the switch by adding a function for recognizing a command voice to the device with speech recognition function 130. In this case, a character string is displayed on the device with speech recognition function 130. It is necessary to add a function for determining whether the voice is an editing command or an editing command.

  Further, when the device 130 with the voice recognition function has a function of storing a character string, the result converted into the character string can be accumulated on the spot. With this configuration, it is possible to record the content uttered with a smaller storage capacity than that for storing voice. Moreover, since it is converted into a character string, searching and the like are facilitated. When the decoded speech is stored in combination with a character string, the usefulness is further increased. Specifically, it is possible to search for a character string using a search character string and reproduce a sound corresponding to the searched character string.

  In addition, when the speech recognition engine 150 of the device 130 with a recognition function uses a word speech recognition technology, it is possible to operate the device 130 with a speech recognition function using the recognition result. For example, when the apparatus with a voice recognition function is a personal computer and application software is activated on the personal computer, the operation of the application can be performed by voice.

(Seventh embodiment)
FIG. 31 shows an audio processing system according to the seventh embodiment of the present invention. This system includes a headset 170 with a wireless communication function, a device 200 with a voice recognition function as a first device, and a second device (not shown) capable of wireless communication with the device 200 with a wireless function. The device 200 with speech recognition function includes a recognition result transmission unit 230 in addition to the speech reception unit 210 and the speech recognition engine 220, and wirelessly transmits the recognition result to the second device.

  The voice receiving means 210 is the same as the voice receiving means 140 in FIG. The speech recognition engine 220 may use either word speech recognition technology or large vocabulary sentence speech recognition technology. Here, word speech recognition technology is used.

  FIG. 32 shows a configuration of the speech recognition engine 220 when the word speech technology is used. The acoustic analysis unit 223, the model matching unit 225, the recognition vocabulary storage unit 227, and the speech model creation / storage unit 229 are the same as those used in the speech recognition unit provided in the headset 10 with the wireless communication function of the first embodiment. It is the composition.

  The word ID output as a recognition result from the speech recognition engine 220 is input to the recognition result transmission unit 230. The recognition result transmission unit 230 transmits the received word ID to another device. As a transmission method to other devices, wireless communication, wired communication, and the like can be considered, but the means is not limited here.

  FIG. 33 illustrates a specific operation of the speech processing system of FIG. A user wearing the headset 170 with the wireless communication function performs voice control of the air conditioner as the second device via the personal computer with the voice recognition function as the first device.

  The user selects the audio transmission mode with the function selection switch 174 of the headset. Therefore, the voice “turn on the air conditioner” detected by the microphone 173 is encoded by the voice transmission means 183 and transferred to the personal computer by wireless communication.

  FIG. 34 shows an example of stored contents of the recognized vocabulary storage unit 227 in the personal computer. The recognition vocabulary storage unit 227 corresponds to the vocabulary of “Eakontsu”, “Eatontotume”, “Ondo Rae”, and “Ondo Sageru”, respectively, with word IDs “01”, “02”, “03” and “04” are given and stored. When the personal computer recognizes the vocabulary “Eakontsu”, the word ID “01” is wirelessly transmitted to the air conditioner.

  According to the stored contents of the recognized vocabulary storage unit 227, new stored contents are created and stored in the speech model creation / storage unit 229. In this example, acoustic models corresponding to the words “Eakontsu,” “Eakontome,” “Odonageru,” and “Ondosageru” are created and paired with the word ID of each word. Will be remembered.

  On the other hand, as shown in FIG. 35, the air conditioner stores each word ID and the corresponding operation as a set, and performs an operation corresponding to the word ID when a specific word ID is received.

  The coded speech received by the speech receiving means 210 of the personal computer is converted into a speech signal by the coded speech decoding unit and input to the speech recognition engine 220. The audio signal is converted into a characteristic parameter series by the acoustic analysis unit 223 and input to the model matching unit 225. The model collation unit 225 collates the input feature parameter series with the speech model of each word stored in the acoustic model creation / storage unit 229. When the similarity of the speech model corresponding to “Eakontsu” is the highest, the collation unit 225 outputs the word ID “01” as the recognition result.

  The word ID “01” is input to the recognition result transmission unit 230, and the word ID “01” is transmitted to the air conditioner by wireless communication. When the air conditioner receives the word ID “01”, the operation of the air conditioner function corresponding to the word ID is started according to the table of FIG.

  With this configuration, the user's voice detected by the microphone 173 of the headset 170 with the wireless communication function can be recognized by the apparatus 200 with the voice recognition function almost in real time, and the recognition result can be transmitted to another device. It becomes.

  When the device 200 with a speech recognition function has a large computing capacity like a personal computer, the speech recognition engine 220 has fewer functional restrictions than the speech recognition unit 177 of the headset, for example, greatly increases the recognition vocabulary. Can be increased. Even if the voice recognition function of the device with voice recognition function 200 cannot be used for some reason, if the function selection switch 174 is switched so as to be processed by the voice recognition unit 177 of the headset, device operation using voice can be performed. It is possible to continue.

  When the large vocabulary speech recognition technology is used for the speech recognition engine 220 as in the speech recognition engine 150 of FIG. 24, the result of conversion into a character string can be immediately transferred to another device. Since the communication amount necessary for transferring the character string is smaller than the communication amount necessary for transferring the voice, the communication amount can be reduced. This system can recognize the voice almost simultaneously with the utterance. In the conventional technology that recognizes the accumulated speech and transfers the result, the speech recognition technology is used after all the utterances are finished, and then the transfer is performed. In the system of the sixth embodiment, since the voice is recognized in parallel with the user's utterance, the time delay can be reduced.

  As described above, in the above-described embodiment, the word recognition is described as an example of the voice recognition in the headset or on the external device side, but the present invention is not limited to this. In particular, in the headset, any speech recognition can be performed as long as it is a simple speech recognition with a small amount of calculation, memory, and power consumption, such as continuous word recognition, sentence recognition, word spotting, and speech intent understanding.

1 is a schematic diagram of a headset with a wireless communication function according to a first embodiment of the present invention. FIG. 2 is a schematic block diagram of the headset of FIG. It is a figure which shows an example of the function selection switch of FIG. It is a figure which shows the internal structural example of the speech recognition part of FIG. It is a figure which shows the example of the memory content of the recognition vocabulary memory | storage part of FIG. It is a figure which shows the response | compatibility of word ID which the air conditioner received, and the operation | movement of an air conditioner. It is a figure which shows ON / OFF control of voice recognition mode by a function selection switch. It is the schematic which shows the system configuration | structure of the headset with a wireless function which concerns on 2nd Embodiment of this invention. It is a figure which shows an example of the function selection switch used with the headset of FIG. FIG. 9 is a diagram showing an internal configuration of the voice transmission means shown in FIG. 8. It is a figure which switches and selects voice recognition and voice transmission processing by a function selection switch. It is the schematic which shows the system configuration | structure of the headset with a wireless communication function which concerns on 3rd Embodiment of this invention. It is a figure which shows an example of the function selection switch shown in FIG. It is a figure when the voice recognition mode or the voice transmission mode is selected by the function selection switch of FIG. FIG. 14 is a diagram illustrating an example in which neither voice recognition nor voice transmission is performed in the OFF mode by the function selection switch of FIG. 13. It is the schematic which shows the system configuration | structure of the headset with a wireless communication function which concerns on 4th Embodiment of this invention. It is a figure which shows an example of the function selection switch of FIG. It is a figure when the voice recognition mode or the voice transmission mode is selected with the function selection switch of FIG. It is a figure which shows the example which processes an audio | voice by both voice recognition and audio | voice transmission with the function selection switch of FIG. It is the schematic which shows the system configuration | structure of the headset with a wireless communication function which concerns on 5th Embodiment of this invention. It is a figure which shows an example of the function selection switch of FIG. It is a figure when the voice recognition mode or the voice transmission mode is selected by the function selection switch of FIG. FIG. 18 is a diagram when the function selection switch in FIG. 17 selects a mode for processing both voice recognition and voice transmission, or an OFF mode in which neither is processed. It is a schematic block diagram of the speech processing system which concerns on 6th Embodiment of this invention. It is a figure which shows the structural example of the audio | voice reception means of the apparatus with a speech recognition function in the system of FIG. It is a figure which shows the structural example of the speech recognition engine of the apparatus with a speech recognition function in the system of FIG. It is a figure which shows the usage example of the system of FIG. It is a figure which shows the example of a memory content of the recognition vocabulary memory | storage part of FIG. It is a figure which shows the example of a memory content of the language model memory | storage part of FIG. It is a figure which shows the example of a screen display of the apparatus with a speech recognition function of FIG. It is a figure which shows the modification of the speech processing system which concerns on 6th Embodiment of this invention. FIG. 32 is a configuration example of a speech recognition engine of a device with a speech recognition function in the system of FIG. 31. FIG. It is a figure which shows the usage example of the audio processing system shown in FIG. It is a figure which shows the example of the memory content of the recognition vocabulary memory | storage part in the system of FIG. It is a figure which shows the response | compatibility of word ID which the air conditioner received via PC in the usage example shown in FIG. 33, and operation | movement of an air conditioner.

Explanation of symbols

10, 110, 170 Headset 13, 113, 173 Microphone 14, 51, 61, 71, 81, 114, 174 Function selection switch 17 Speaker 16 CPU board 17 Wireless communication module 19, 119, 181 Function selection unit 20, 50, 60, 70, 80 Function selection means 21, 121, 75 A / D converters 23, 123, 177 Speech recognition units 25, 125, 178, 230 Recognition result transmission means 41 Recognition signal blockers 43, 151, 223 Acoustic analysis 45, 155, 225 Model matching unit
47, 157, 227 Recognition vocabulary storage unit 49, 159, 229 Speech model creation / storage unit 53, 153, 183 Speech transmission means 55 Transmission signal blocker 57 Speech coding unit 59 Speech transmission unit 130, 200 With speech recognition function Devices 140 and 210 Speech receiving means 141 Encoded speech receiving unit 143 Encoded speech decoding unit 150 and 220 Speech recognition engine 161 Language model storage unit 163 Word ID notation conversion

Claims (8)

A headset with wireless communication function;
An external device capable of wireless communication with the headset, and the headset with the wireless communication function includes:
A microphone that detects a voice of a wearer of the headset and generates an audio signal;
Voice recognition means for recognizing the voice signal and generating an identification signal corresponding to the content of the recognized voice signal;
A speech processing system comprising: recognition result transmission means for sending the identification signal generated by the voice recognition means to the external device by wireless communication, wherein the external device performs an operation corresponding to the received identification signal.   The audio processing system according to claim 1, wherein the external device includes a table that stores a plurality of identification signals and operations corresponding to the identification signals in association with each other.   The voice processing system according to claim 1, wherein the headset further includes a function selection unit that switches whether or not a voice signal generated by the microphone is processed by a voice recognition unit. A headset with wireless communication function;
An external device with a voice recognition function capable of wireless communication with the headset, the headset with a wireless communication function,
A microphone that detects a voice of a wearer of the headset and generates an audio signal;
Audio transmission means for transmitting the audio signal to the external device by wireless communication, the external device,
Audio receiving means for receiving an audio signal transmitted from the headset;
A speech processing system comprising speech recognition means for recognizing the received speech signal.   The voice processing system according to claim 4, wherein the external device performs an operation according to a recognition result by the voice recognition unit. The external device further includes a display unit,
The voice recognition means recognizes the received voice signal, generates an identification signal corresponding to the content of the recognized voice signal, converts the identification signal into a character, and outputs it.
The voice processing system according to claim 4, wherein the display unit displays a character that is a recognition result.   The voice processing system according to claim 4, wherein the headset further includes voice recognition means for recognizing the voice signal. A headset with wireless communication function;
A first external device having a voice recognition function and capable of wireless communication with the headset;
A second external device capable of wireless communication with the first external device, and the headset with the wireless communication function includes:
A microphone that detects a voice of a wearer of the headset and generates an audio signal;
Audio transmission means for transmitting the audio signal to the first external device by wireless communication, and the first external device comprises:
Audio receiving means for receiving an audio signal transmitted from the headset;
Voice recognition means for recognizing the received voice signal and generating an identification signal corresponding to the content of the recognized voice signal;
Recognition result transmission means for transmitting the identification signal to the second external device by wireless communication, wherein the second external device performs an operation corresponding to the word ID received from the first external device. Processing system.

Images