JP4299768B2 - Voice recognition device, method, and portable information terminal device using voice recognition method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To prevent the degradation in a recognition rate by amplifying a speech level inputted according to the use state of a mobile information terminal into an appropriate speech level. <P>SOLUTION: The speech recognition device for recognizing the speech inputted to a microphone is equipped with: an amplifier 108 which amplifies the speech signal outputted from the microphone; a speech level detector 103A which detects the amplified speech level; a transmission gain information storage section 106A which memorizes a transmission gain, an appropriate speech level and a time constant for updating the transmission level; a transmission gain setting controller 103B which reads out the transmission gain, the appropriate speech level, and the time constant, updates the transmission gain by adding the value obtained by multiplying the gain ought to adjust the detected speech level to the appropriate speech level by the time constant and memorizes the updated transmission gain in the transmission gain information storage section; and a speech recognition section 111 which is inputted with the updated transmission gain and performs the speech recognition. <P>COPYRIGHT: (C)2006,JPO&amp;NCIPI

Description

  The present invention relates to a speech recognition apparatus used in a movable state. In particular, the present invention provides a speech recognition apparatus, method, and speech recognition method that can prevent a reduction in recognition rate due to an inability to amplify an input speech level according to a use state at the time of transmission to an appropriate speech level. The present invention relates to the portable information terminal device used.

In recent years, mobile phones, which are mobile devices, have been equipped with a voice recognition function. In mobile phones, voice is input from the microphone of the transmitter, amplified to an appropriate voice level, voice recognition is performed, and the recognition result is displayed on the display. And sounding from the speaker.
At the time of voice recognition, the highest recognition performance can be obtained when a voice signal having an appropriate voice level is input, but the recognition rate is lowered regardless of whether it is lower or higher than the appropriate voice level.

Thus, in a mobile phone having a voice recognition function, a standard transmission gain is stored and set in the microphone amplifier of the transmission unit as follows.
FIG. 16 is a diagram for explaining an example of use of voice recognition in a mobile phone which is a premise of the present invention.
As shown in FIG. 5A, voice recognition is performed in a manner of holding a call state in which a speaker of a receiving unit in a mobile phone is in close contact with a user's ear.

In this case, the standard distance d1 between the microphone of the transmitter of the mobile phone and the user's mouth, based on the length and shape of the mobile phone and the standard size of the user's head, and the standard loudness of the voice Under such conditions, a standard transmission gain Ga is set in the microphone amplifier of the transmission unit.
However, in practice, the user does not secure the standard distance d1, which is the same usage condition, in terms of how to hold the cellular phone, the size of the head, the volume of the voice, etc. Since the user's usage conditions vary, the voice level input to the microphone of the transmitter varies, and is amplified with the standard transmission gain Ga set in the amplifier. Therefore, there is a problem in that the voice level that is set does not become an appropriate voice level, and thus the voice recognition rate is lowered and a malfunction occurs.

As shown in FIG. 5B, voice recognition is performed in a state where a telephone call is made while the receiving unit of the mobile phone is separated from the user's ear and the display unit of the mobile phone is viewed.
In this case, it is standard for the microphone of the microphone of the transmitter so that the standard loudness becomes an appropriate voice level at a standard distance d2 between the microphone of the microphone of the portable phone and the user's mouth. A simple transmission gain Gb is set.

  However, in practice, the standard distance d2 between the microphone of the transmitter in the mobile phone and the user's mouth cannot be ensured, resulting in a variation in distance, and fluctuations without speaking with the standard sound volume. Since the distance is larger than in the case of FIG. 16 (a), it is easily affected by noise, so that the voice input to the microphone fluctuates, and the voice level amplified by the standard transmission gain Gb set in the amplifier is There is a problem in that an appropriate voice level is not achieved and the voice recognition rate is lowered.

  Further, in the method of use in which the mobile phone is in close contact with the user's ear as shown in FIG. 5A, the method of use in which the mobile phone is separated from the user's ear as shown in FIG. Since the distances d1 and d2 between the user's mouths are greatly different from each other, standard transmission gains Ga and Gb set in the microphone amplifier of the transmission unit are different from each other. The standard transmission gain Ga set in the microphone amplifier of the transmission unit in the mobile phone shown in FIG. The setting change must be made to the standard transmission gain Gb set in the microphone amplifier of the transmission unit in the cellular phone shown in FIG. If this setting is not changed, the speech recognition rate is significantly reduced.

  The reverse is also true, and the standard set in the microphone amplifier of the transmitter in the cellular phone shown in FIG. (B) with respect to the change in usage from FIG. (B) to FIG. (A). The transmission gain Gb must be changed to the standard transmission gain Ga set in the microphone amplifier of the transmission unit in the cellular phone shown in FIG. If this setting is not changed, the speech recognition rate is significantly reduced.

For this reason, with regard to the switching of the usage method from (a) to (b) in this figure and the switching from the usage method to (a) in this figure from (b) to (a), the standard of the microphone amplifier is changed at each switching. The settings of the transmission gains Ga and Gb must be changed, and the operation becomes complicated.
In other words, the voice recognition function which is the premise of the present invention has a problem that the voice recognition rate is lowered because the volume of the user of the mobile phone and the usage method do not necessarily match the assumed conditions. is there.

  Conventionally, in order to automatically adjust the reception volume during a call, the level detector detects and calculates the DC component power value A1 of the transmitted voice signal, and the subtractor writes the transmitted power value A1 and the memory. Amplification amount A2 is calculated by subtracting the reference power value A0, and temporarily stored in the memory unit. At the next timing, the received audio signal is input to the amplifier in the signal comparison unit via the audio signal processing unit. When this is done, the amplification amount A2 of the received voice read from the memory unit is read out, and the gain of the amplifier is variably controlled to amplify the received voice signal by A2 times, thereby variably adjusting the received volume according to the transmission volume. There is something that can be done (for example, see Patent Document 1).

  However, in Patent Document 1, in order to adjust the reception volume, the amount of amplification is calculated, the gain of the amplifier is variably controlled, and the received audio signal is amplified by the calculated amount of amplification. When used in close contact with the ears, the sound level amplified by the microphone amplifier of the transmitter is not an appropriate sound level due to variations in the size of the head and the user's sound level. When the telephone is used away from the user's ear, the voice level amplified by the microphone amplifier of the transmitter is appropriate due to variations in the distance between the microphone and mouth of the transmitter and noise. If the usage method is changed from a state in which the mobile phone is in close contact with the ear to a state in which the mobile phone is separated from the ear, or If the usage method is changed from a state where the mobile phone is separated from the ear to a state where the mobile phone is in close contact with the ear, the standard transmission gain that should be set for the microphone amplifier of the transmitter is different. It is necessary to change the setting, and the problem that the operation becomes complicated cannot be solved.

  Conventionally, the quality of the listening sound is improved by outputting the voice of a specific conference participant whose voice is low or speaking away from the microphone so as to be the same level as the voice output level of other conference participants. Therefore, in a conference telephone device that performs a hands-free remote conference using a plurality of microphones and speakers, a speech recognition circuit that decomposes a speech signal input from one or more microphones into speech elements, a memory circuit, Means for storing in the memory circuit the output of the voice recognition circuit of the specific talker inputted from the one or more microphones, and the output of the voice recognition circuit during the remote conference and the contents stored in the memory circuit. A specific conversation using a matching circuit for matching, one or more gain setting circuits provided for one or more microphones, and a matching circuit There are those provided with means for increasing control the gain of the microphone corresponding gain setting circuit voice of a particular conversation who is input when the voice recognition (for example, see Patent Document 2).

  However, in the above-mentioned patent document 2, the voice of a specific conference participant whose voice is low or speaks away from the microphone is output so as to be the same level as the voice output level of other conference participants. Thus, when the mobile phone is used in close contact with the user's ear, the sound level amplified by the microphone amplifier of the transmitter is appropriate due to the variation in the head size and the variation in the user's voice level. When the mobile phone is used away from the user's ear, the voice level is not amplified, and it is amplified by the microphone of the microphone of the transmitter due to variations in the distance between the microphone of the microphone and the mouth and noise. The problem that the voice level does not reach the proper level and the voice recognition rate is reduced cannot be solved, and the mobile phone is moved away from the ear when it is in close contact with the ear. When the usage method is changed to a normal state, or conversely, when the usage method is changed from a state where the mobile phone is away from the ear to a state where it is in close contact with the ear, the standard transmission to be set for the microphone amplifier of the transmitter section Since the speech gains are different, it is necessary to change the standard transmission gain setting, and the problem that the operation becomes complicated cannot be solved.

  In addition, conventionally, in order to obtain a hand-free car phone device that can automatically adjust the volume level even if the amount of external noise changes during a hands-free call with a car phone and can perform a clear dialogue, A hand-free call circuit that supplies conversational voice to this radio, a microphone that inputs conversational voice to this hand-free call circuit, a speaker that outputs conversational voice from the hand-free call circuit, and the other party's name are input by voice In a hands-free car telephone device having a voice recognition device that performs voice recognition processing and instructs the radio device to make a call according to a recognition result, according to noise data detected by the voice recognition device during a hand-free call. Automatically adjust at least one of the input voice level input from the microphone and the volume of the conversation voice output from the speaker Also that was Unishi (e.g., see Patent Document 3).

  However, in the above-mentioned Patent Document 3, the volume level can be automatically adjusted even in the case of a hand-free call in a car phone even if the level of external noise changes, and a clear dialogue can be performed. In addition, when the mobile phone is used in close contact with the user's ear, the sound level amplified by the microphone amplifier of the transmitting unit due to the variation in the size of the head and the sound level of the user is an appropriate sound level. Furthermore, when the mobile phone is used away from the user's ear, the voice level amplified by the microphone amplifier of the transmitter due to variations in the distance between the microphone of the receiver and the mouth and noise Cannot solve the problem that the voice recognition rate declines due to an inappropriate voice level, and the mobile phone is used close to the ear and away from the ear. If the usage method is changed from a state where the mobile phone is separated from the ear to a state where the mobile phone is in close contact with the ear, the standard transmission gain to be set in the microphone amplifier of the transmission unit differs. Therefore, it is necessary to change the standard transmission gain setting, and the problem that the operation becomes complicated cannot be solved.

  Also, conventionally, it is possible to correctly detect the voice section of the input telephone voice without incurring the beginning, ending or saturation of the input telephone voice regardless of the change in the power level of the input telephone voice. For the purpose of improvement, a preamplifier for giving a predetermined amplification gain to telephone voice input via a telephone line, and a voice section of the telephone voice input via the preamplifier are detected. A voice recognition unit for detecting a feature of the telephone voice and recognizing the telephone voice, a voice response part for sending a predetermined response voice to the telephone line according to the voice recognition result, and a process of the voice recognition and voice response Means for detecting the power level of the first telephone voice input via the telephone line, and setting the amplification gain in the amplifier according to the detected power level A voice input device having a that means (e.g., see Patent Document 4).

  However, in Patent Document 4, the power level of the telephone voice is detected, the amplification gain in the amplifier is set according to the detected power level, the voice section of the input telephone voice is correctly detected, and the recognition performance for the input telephone voice is improved. However, as described above, when the mobile phone is used in close contact with the user's ear, it is amplified by the microphone amplifier of the transmitter due to the variation in the head size and the variation in the voice level of the user. The voice level does not reach an appropriate level, and when the mobile phone is used away from the user's ear, the distance between the microphone of the transmitter and the mouth and the noise of the microphone of the transmitter are reduced due to noise. The problem that the voice level amplified by the amplifier does not reach the proper voice level and the voice recognition rate is reduced cannot be solved. When the usage method is changed from being in close contact with the ear to being away from the ear, or vice versa, when the usage method is changed from being away from the ear to being in close contact with the ear, Since the standard transmission gain to be set in the amplifier is different, it is necessary to change the standard transmission gain setting, and the problem that the operation becomes complicated cannot be solved.

  Conventionally, in an automobile telephone device with an emergency call function, in order to reliably transmit the user's transmitted voice to the emergency call center, a microcomputer is sent from the passenger in the event of an emergency such as a traffic accident. When it is determined that the talk level is less than the predetermined value, the transmission gain is increased from the normal state and the output signal from the microphone is amplified. Therefore, the gain control amplifier automatically performs the normal state. The output signal can be output at a higher power level, so that the uplink communication signal can be sent to the base station at a higher power level than in the normal state. There is something that can be communicated to the operator (see, for example, Patent Document 5).

  However, in the above-mentioned Patent Document 5, when an emergency such as a traffic accident occurs, a process for increasing the transmission gain from the normal state and amplifying the output signal from the microphone is performed to ensure the transmission voice of the passenger. As described above, when the mobile phone is used in close contact with the user's ear, as described above, the microphone of the transmitter is caused by the variation in the size of the head and the variation in the voice level of the user. When the mobile phone is used away from the user's ears, the sound level amplified by the amplifier in (2) is not appropriate. We cannot solve the problem that the voice level amplified by the microphone of the talking part does not reach the proper voice level and the voice recognition rate decreases. When the usage method is changed from a state where the machine is in close contact with the ear to a state where it is separated from the ear, or conversely, when the usage method is changed from a state where the mobile phone is separated from the ear to a state where the mobile phone is in close contact with the ear, Since the standard transmission gain to be set in the amplifier of the microphone is different, it is necessary to change the standard transmission gain setting, and the problem that the operation becomes complicated cannot be solved.

Japanese Patent Laid-Open No. 11-239093 JP 61-161863 A JP-A-4-261254 Japanese Patent Laid-Open No. 1-142799 JP 2004-80697 A

  Therefore, in view of the above problems, the present invention amplifies a voice level input according to a usage state of a portable information terminal device to an appropriate voice level, and enables a speech recognition apparatus and method capable of preventing a reduction in recognition rate. It is another object of the present invention to provide a portable information terminal device using the voice recognition method.

In order to solve the above problems, the present invention provides a speech recognition apparatus for recognizing speech input to a microphone of a transmitter, and an amplifier that amplifies a speech signal output from the microphone of the transmitter by a transmission gain. A voice level detector for detecting the voice level amplified by the amplifier, and an initial value of the transmission gain, a transmission gain, an appropriate voice level, and a transmission gain update time constant for updating the transmission gain. A transmission gain information storage unit, a transmission gain, an appropriate voice level, a transmission gain update time constant are read from the transmission gain information storage unit, the transmission gain is set in the amplifier, and the voice level detection is performed. the value obtained by multiplying the time constant for the transmission gain update the gain to be within the appropriate voice level detected speech level in parts is added to the transmission gain updates the transmission gain, A transmission gain setting controller configured to store the transmission gain new was the transmission gain information storage unit, and a speech recognition unit which performs speech recognition by entering the voice signal amplified by the amplifier, feeding by keystroke The speech gain update time constant can be adjusted, the speech recognition unit can process speech recognition for a plurality of test utterances, and the transmission gain setting control unit can obtain the optimum value of the transmission gain in advance. And a test unit which stores the transmission gain information storage unit as an initial value of the transmission gain.

Further, the transmission gain setting control unit reads an initial value of the transmission gain from the transmission gain information storage unit and sets it in the amplifier during the first utterance during speech recognition.
Further, the transmission gain setting control unit, when the updated transmission gain is stored in the transmission gain information storage unit, the transmission gain at the first utterance when the speech recognition is resumed. The updated transmission gain is read from the information storage unit and set in the amplifier.

  Further, the transmission gain setting control unit reads the initial value of the transmission gain from the transmission gain information storage unit at the time of the first utterance at the time of speech recognition, sets the initial value of the transmission gain in the amplifier, and performs speech recognition on the speech recognition unit. If the activation of speech recognition is detected within a predetermined time after the speech recognition is confirmed, the initial value of the transmission gain is read from the transmission gain information storage unit, set in the amplifier, and the acquired speech Based on the level information, the initial value of the transmission gain is updated and stored in the transmission gain information storage unit, or the transmission gain before update is read from the transmission gain information storage unit, set in the amplifier, and acquired. The transmission gain is updated based on voice level information, stored in the transmission gain information storage unit, and the updated transmission gain at the next utterance is read from the transmission gain information storage unit and set in the amplifier. And, if the activation of the voice recognition is not detected within the predetermined time to end the speech recognition process.

Further, the transmission gain setting control unit determines whether the voice level detected by the transmission gain information storage unit is within a certain range centered on the appropriate voice level or outside the certain range. Then, the time constant when it is within the certain range is made smaller than the time constant when it is outside the certain range.
Further, the transmission gain Gn updated by the transmission gain setting control unit is expressed by the following equation: Gn = Gn-1−K × 20 × log (Xn / C) dB
(K: Time constant for transmission gain update for transmission gain update (0.0
<K ≦ 1.0),
n: Number of voice recognition (= 1, 2, 3 ...),
C: Appropriate audio level,
Xn: voice level detected by the voice level detector
It is represented by

Et al is, the present invention provides a speech recognition method for recognizing a speech input to the microphone of the mouthpiece unit, appropriate sound level, a step of storing the transmission gain, the audio signal output from the microphone of the transmitter section The transmission gain is a value obtained by multiplying a gain that should make the detected voice level the appropriate voice level and a time constant. The step of adding and updating the transmission gain and storing the updated transmission gain, the step of inputting the amplified speech signal and performing speech recognition, and the input of the amplified speech signal and performing speech recognition The process and the time constant for updating the transmission gain by key operation can be adjusted, voice recognition is processed for multiple test utterances, the optimal value of the transmission gain is obtained in advance, and the initial value of the transmission gain Work A speech recognition method comprising:

Furthermore, the present invention provides a portable information terminal device using a voice recognition method, wherein, in addition to the portable information terminal function of the portable information terminal device, the voice level input according to the usage state of the portable information terminal device. A voice recognition function for performing voice recognition by amplifying to an appropriate voice level is provided.
Furthermore, the recognition result recognized by voice is sounded on the speaker of the earpiece of the portable information terminal device and displayed on the display unit of the portable information terminal device.

  As described above, according to the present invention, an audio signal output from the microphone of the transmission unit is input with the audio signal amplified by the transmission gain to perform audio recognition, and the amplified audio signal The voice level is detected, and the transmission gain is updated by adding the value obtained by multiplying the detected voice level to the appropriate voice level by the time constant to the transmission gain, and the updated transmission gain is stored. Since it did in this way, especially the audio | voice level input according to the use condition of the transmission part of a portable information terminal device is amplified to an appropriate audio | voice level, and it becomes possible to prevent a recognition rate fall.

  When the portable information terminal device is used in close contact with the user's ear, the sound level amplified by the microphone amplifier of the transmitter is appropriate due to the variation in the head size and the variation in the user's voice level. Furthermore, when the portable information terminal device is used away from the user's ear, the sound level amplified by the microphone amplifier of the transmitter due to the variation in the distance between the microphone of the receiver and the mouth Becomes an appropriate voice level, the speech recognition rate is improved, and the portable information terminal device is changed from the state of being in close contact with the ear to the state of being separated from the ear, or vice versa. When the usage method is changed from a state in which the terminal device is away from the ear to a state in which the terminal device is in close contact with the ear, the transmission gain setting is automatically changed, and it is necessary to change the transmission gain setting as before. Eliminated, operation is simplified.

Hereinafter, embodiments of the present invention will be described with reference to the drawings.
FIG. 1 is a block diagram showing a schematic configuration of a portable information terminal device according to the present invention. As shown in the figure, an antenna 101 is provided in a portable information terminal device 100 that is a mobile device, and the antenna 101 performs wireless communication with a base station (not shown).
A radio unit 102 is connected to the antenna 101, and the radio unit 102 modulates a transmission signal to the antenna 101 and demodulates a reception signal from the antenna 101.

A control unit 103 is connected to the wireless unit 102, and the control unit 103 controls the entire portable information terminal device 100 including the wireless unit 102.
An operation unit 104 is connected to the control unit 103. The operation unit 104 includes a CPU (Central Processing Unit), and performs operations such as operation of a mobile phone and start of voice recognition. When the voice recognition start key is pressed by the operation unit 104, the control unit 103 transmits a voice recognition start command to the voice recognition unit 111 described later.

Further, a display unit 105 is connected to the control unit 103, and the display unit 105 displays numbers, characters, images, speech recognition recognition results, and the like.
Further, a memory 106 is connected to the control unit 103, and the memory 106 can rewrite data and stores various information for controlling the portable information terminal device 100 and transmission / reception data.
Further, an A / D / D / A converter 107 is connected to the control unit 103, and the A / D / D / A converter 107 converts an audio signal of a transmission sound to the control unit 103 from an analog signal to a digital signal. The voice signal of the received sound from the control unit 103 is converted from a digital signal to an analog signal.

An amplifier (amplifier) 108 is connected to the A / D / D / A converter 107, and the amplifier 108 adjusts sensitivity by changing the amplification factor when the transmission gain is set by a transmission gain setting control unit 103B described later. Then, an audio signal which is an analog signal from a microphone 109 described later is amplified.
The amplifier 108 is connected to the microphone 109 of the transmission unit. The microphone 109 receives the user's transmission sound, converts it into an electrical signal, and outputs it to the amplifier 108 as an analog audio signal.

The A / D / D / A converter 107 is connected to the speaker 110 of the receiver. The speaker 110 receives an electrical signal of the analog signal of the received sound from the A / D / D / A converter 107 and converts the electrical signal into the received sound. It converts and rings the received sound, especially the recognition result of voice recognition.
A speech recognition unit 111 is connected to the control unit 103, and the speech recognition unit 111 is composed of a DSP (Digital Signal Processor) LSI (Large Scale Integrated Circuit). The speech signal from the microphone 109 is amplified by the amplifier 108, and the A / A Voice recognition processing is performed on the voice data digitized by the D / D / A converter 107 and input via the control unit 103, and the voice recognition recognition result is displayed on the display unit 105 via the control unit 103. The start sound at the time of voice recognition and the recognition result of voice recognition are caused to sound on the speaker 110 by voice.

The memory 106 is provided with a transmission gain information storage unit 106A. The transmission gain information storage unit 106A is a transmission gain initial value (dB) as various information for optimizing the transmission gain to the amplifier 108 during speech recognition. The updated transmission gain, the appropriate voice level for speech recognition, the time constant for updating the transmission gain, and the like are stored.
The control unit 103 is provided with a voice level detection unit 103A, and the voice level detection unit 103A always amplifies a voice signal from the microphone 109 of the transmission unit by the amplifier 108 at the time of voice recognition, and A / D / D / A After being converted into voice data by the converter 107 and input to the control unit 103, a voice section is detected from the voice data to detect the voice level of the voice signal of the transmission sound.

  The control unit 103 is provided with a transmission gain setting control unit 103B. The transmission gain setting control unit 103B uses various types of information held in the transmission gain information storage unit 106A during speech recognition, and further transmits a transmission gain. Using the voice level detected by the setting control unit 103B, the gain obtained from the voice level and the appropriate voice level for the next voice recognition is multiplied by the time constant and added to the transmission gain obtained at the previous voice recognition. Update the value as the new transmission gain value.

Furthermore, the transmission gain setting control unit 103B sets the transmission gain obtained in the previous speech recognition in the amplifier 108 in the next speech recognition.
The transmission gain update formula performed by the transmission gain setting control unit 103B described above will be described below.
The voice level input to the microphone 109 is An, the voice level detected by the voice level detector 103A is Xn, the appropriate voice level is C, and the transmission gain Gn update formula is expressed by the following formula.

Gn = Gn-1 -K * 20 * log (Xn / C) dB
... (1)
(K: Time constant for updating transmission gain (0.0 <K ≦ 1.0),
n: Number of voice recognition (= 1, 2, 3 ...))
Xn = An * 10 ^{Gn-1 / 20} (2)

  Thus, when the next speech recognition is activated, the transmission gain reflecting the speech level of the previous speech recognition result is set in the amplifier 108. Therefore, the user's voice volume and usage method are assumed. Depending on the condition, a sound level suitable for speech recognition can be obtained.

  FIG. 2 is a diagram for explaining various examples of information for optimizing the transmission gain at the time of speech recognition, held in the transmission gain information storage unit 106A in FIG.

  As shown in this figure, in the transmission gain information storage unit 106A, the distance d1 (see FIG. 16A) between the user's mouth and the microphone 109 of the transmission unit in the portable information terminal device 100 is 3 cm. If there is, or if the distance d2 (see FIG. 16 (b)) is 7 cm, one is selected and “0.00 dB” is held as the initial value G0 of the transmission gain as a reference, and further updated Gain (dB) is maintained. This updated transmission gain is used as a value set in the amplifier 108 at the next speech recognition.

Further, the transmission gain information storage unit 106A stores a proper speech level C for speech recognition, and for example, “1000” is held as the proper speech level C, and a time constant K for updating the transmission gain is stored. A plurality of “1.0”, “0.5”,... Are held so as to be selectable as the constant K.
FIG. 3 is a flowchart for explaining a series of operation examples of transmission gain update processing of the transmission gain setting control unit 103B in FIG.
As shown in the figure, in step 201, the control unit 103 detects that the voice recognition start key is pressed by the operation unit 104 and detects that voice recognition is activated.

In step 202, the transmission gain setting control unit 103B detects the activation of the control unit 103, reads the transmission gain Gn-1 before update from the transmission gain information storage unit 106A, and sets it in the amplifier 108. When the transmission gain before update is not held in the transmission gain setting control unit 103B, the initial value of the transmission gain is set in the amplifier 108.
In step 203, after the transmission gain setting control unit 103B sets the transmission gain Gn-1 before update in the amplifier 108, the control unit 103 uses the amplifier 108 to recognize the voice signal input from the microphone 109 for speech recognition. For the input voice adjusted to the voice level (see Expression (2)), the voice recognition unit 111 is activated to perform voice recognition.

In step 204, the transmission gain setting control unit 103B waits for confirmation of the recognition result from the speech recognition unit 111.
In step 205, the transmission gain setting control unit 103B acquires the recognition result and the voice level information detected by the voice level detection unit 103A after the recognition result of the voice recognition is confirmed. The recognition result is displayed on the display unit 105, and the speaker 110 is sounded by voice.

In step 206, transmission gain update processing (see equation (1)) is performed.
In step 207, the transmission gain setting control unit 103B stores the updated transmission gain value in the transmission gain information storage unit 106A, and ends the process.
In this way, the transmission gain is held in the transmission gain information storage unit 106A, and the transmission gain held in the transmission gain information storage unit 106A is read and the transmission gain is updated every time speech recognition is performed. This makes it possible to acquire the optimum transmission gain. Thereby, the voice level at the time of voice recognition becomes an appropriate voice level, and the voice level suitable for voice recognition is continuously secured thereafter. In particular, even if a user who does not use the recommended method of voice activation, voice recognition is performed several times, so that the transmission gain is updated according to the user's usage and voice volume. The optimum audio level is used.

When a portable information terminal device including a mobile phone is used in close contact with the user's ear, the voice level amplified by the microphone amplifier of the transmitter due to the variation in head size and the variation in the user's voice level When the portable information terminal device is used away from the user's ear, the microphone of the transmitter is amplified due to variations in the distance between the microphone of the receiver and the mouth and noise. When the voice level amplified in step 1 becomes an appropriate voice level, the voice recognition rate is improved, and the usage method is changed from a state where the portable information terminal device is in close contact with the ear to a state where it is separated from the ear, or Conversely, when the usage method is changed from a state where the portable information terminal device is separated from the ear to a state where the portable information terminal device is in close contact with the ear, the transmission gain setting is automatically changed, and the transmission gay is performed as in the conventional case. There is no need to perform configuration changes, operation is simplified.
A specific example will be described below.

FIG. 4 is a calculation example of the transmission gain by the transmission gain setting control unit 103B in FIG. 1, in which the distance from the microphone 109 to the user's mouth (d1 = 3 cm) is small, the sound level input to the microphone 109 is small, It is a figure explaining the example in case there is no variation and the time constant K = 1.0.
In this figure, as an example, it is assumed that the voice levels input to the microphone 109 during the first, second, third, etc. speech recognition are A1 = 700, A2 = 700, A3 = 700,. The appropriate voice level is set to C = 1000, and the transmission gain is calculated as follows.

In the utterance at the time of the first speech recognition, the initial value of the transmission gain is G0 = 0.00 dB, and is set in the amplifier 108 before utterance. In this case, the speech level X1 detected by the speech level detection unit 103A From the equation (2)
A1 / X1 = 1
In this case,
X1 = A1 = 700
It becomes.

The updated transmission gain G1 is calculated from equation (1):
G1 = G0−1.0 × 20 × log (X1 / 1000) dB
= 0.0-1.0 x 20 x log (700/1000) dB
= 3.10 dB
It becomes.

  In the above example, as shown in FIG. 2, the initial value (G0 = 0.00 dB) of the transmission gain is read from the transmission gain information storage unit 106A, and the calculated transmission gain G1 is the updated transmission gain. Is stored in the transmission gain information storage unit 106A, and is used as the transmission gain G1 before update at the time of the next speech recognition. The same applies hereinafter.

In the utterance at the time of the second speech recognition, the transmission gain before update is G1 = 3.10 dB and is set in the amplifier 108 before utterance. In this case, the speech level X2 detected by the speech level detection unit 103A From the equation (2)
X2 = A2 × 10 ^{G1 / 20}
= 700 × 10 ^{3.10 / 20}
= 1000
It becomes.

The updated transmission gain G2 is calculated from equation (1):
G2 = G1-1.0 * 20 * log (X2 / 1000) dB
= 3.10-1.0 x 20 x log (1000/1000) dB
= 3.10 dB
It becomes.

In the utterance after the third voice recognition, the voice level X3 = 1000 and G3 = 3.10 dB detected by the voice level detection unit 103A similar to the second voice recognition are obtained.
That is, at the time of the first speech recognition, the transmission gain setting control unit 103B sets G0 = 0.0 dB in the amplifier 108 before speaking, the speech level detection unit 103A detects the speech level X1 = 700, and G0 = 0.00 dB is updated to G1 = 3.10 dB.

  In this case, since the time constant is K = 1.0 and the voice level A1 = A2 = A3 =... = 700 is inputted to the microphone 109 and there is no variation, the voice level detection unit 103A performs the second voice recognition. The detected voice level X2 is X2 = 1000, which matches the appropriate voice level C = 1000, and the transmission gain update value is G2 = 3.10 dB, which is the optimum value.

In other words, when the distance between the microphone 109 and the user's mouth is constant, the input voice level is constant, and there is no noise in the usage environment, it is preferable to increase the time constant to speed up the follow-up.
In the above example, there is no variation in the sound level input to the microphone 109, but the distance between the microphone 109 and the user's mouth is not constant, the input sound level fluctuates, and is used in a noise environment. In this case, if the time constant is increased, the transmission gain setting changes greatly, and conversely, a non-optimal transmission gain is set.If there is variation, the time constant is reduced as follows. Set the talk gain.

FIG. 5 is a calculation example of the transmission gain by the transmission gain setting control unit 103B in FIG. 1, and the distance from the microphone 109 to the user's mouth (d1 = 3 cm) is small, and the voice level input to the microphone 109 is small. It is a figure explaining an example in case there is variation and time constant K = 1.0.
In this figure, the voice levels input to the microphone 109 during the first, second, third, etc. speech recognition are as follows: A1 = 700, A2 = 750, A3 = 700, A4 = 750, etc. Suppose that other conditions are the same as the above example, the transmission gain is calculated as follows.

In the utterance at the time of the first speech recognition, the initial value of the transmission gain is G0 = 0.00 dB, and is set in the amplifier 108 before utterance. In this case, the speech level X1 detected by the speech level detection unit 103A From the equation (2)
A1 / X1 = 1
In this case,
X1 = A1 = 700
It becomes.

The updated transmission gain G1 is calculated from equation (1):
G1 = G0−1.0 × 20 × log (X1 / 1000) dB
= 0.0-1.0 x 20 x log (700/1000) dB
= 3.10 dB
It becomes.

In the utterance at the time of the second speech recognition, the transmission gain before update is G1 = 3.10 dB and is set in the amplifier 108 before utterance. In this case, the speech level X2 detected by the speech level detection unit 103A From the equation (2)
X2 = A2 × 10 ^{G1 / 20}
= 750 × 10 ^{3.10 / 20}
= 1072
It becomes.

The updated transmission gain G2 is calculated from equation (1):
G2 = G1-1.0 * 20 * log (X2 / 1000) dB
= 3.10-1.0 x 20 x log (1072/1000) dB
= 2.50 dB
It becomes.

In the utterance at the time of the third speech recognition, the transmission gain before update is G2 = 2.50 dB, and is set in the amplifier 108 before utterance. In this case, the speech level X3 detected by the speech level detection unit 103A From the equation (2)
X3 = A3 × 10 ^{G2 / 20}
= 700 × 10 ^{2.50 / 20}
= 933
It becomes.

The updated transmission gain G3 is calculated from equation (1):
G3 = G2-1.0 * 20 * log (X3 / 1000) dB
= 2.5-1.0 * 20 * log (933/1000) dB
= 3.10 dB
It becomes.

In the utterance after the fourth speech recognition, the speech levels X2 = 1072 and X3 = 933 detected by the speech level detector 103A are repeated as in the second and third speech recognition. .
That is, the transmission gain setting control unit 103B sets G0 = 0.0 dB in the amplifier 108 during the first speech recognition, the speech level detection unit 103A detects the speech level X1 = 700, and G0 = 0. Update 00 dB to G1 = 3.10 dB.

  In this case, since the time constant is K = 1.0 and there are variations in the sound levels A1 = 700, A2 = 750, A3 = 700,... Input to the microphone 109, the second, third,. The voice levels X2, X3,... Detected by the voice level detector 103A at the time of voice recognition are X2 = 1072, X3 = 933, and do not coincide with the appropriate voice level C = 1000. 2.50 dB and 3.10 dB, which do not match each other.

As described above, when there is no variation, the appropriate sound level is obtained during the second speech recognition. However, when there is variation, the sound level detected by the sound level detection unit 103A does not become the appropriate sound level.
For this reason, when the user always fluctuates without speaking at the same voice volume under the same conditions, the voice levels A1, A2, A3,... Input to the microphone 109 vary, and the voice level detector 103A Since the detected voice level does not match the appropriate voice level, it is difficult to obtain a high recognition rate during voice recognition.

  Therefore, next, as an example, when the time constant K is set to a value smaller than “1.0”, for example, “0.5” and there is no variation, the audio level detection unit 103A is as follows. The sound level detected by the above is brought close to the appropriate sound level C.

FIG. 6 is an example of calculation of the transmission gain by the transmission gain setting control unit 103B in FIG. 1. The distance from the microphone 109 to the user's mouth (d1 = 3 cm) is small, and the sound level input to the microphone 109 is small. It is a figure explaining an example in case there is no variation and time constant K = 0.5.
In this figure, as an example, the voice levels input to the microphone 109 during the first, second, third, fourth, fifth, etc. speech recognition are as follows: A1 = 700, A2 = 700, With A3 = 700, A4 = 700, A5 = 700, A6 = 700,... And a time constant of 0.5, the transmission gain is calculated as follows.

In the utterance at the time of the first speech recognition, the initial value of the transmission gain is G0 = 0 dB and is set in the amplifier 108 before utterance. In this case, the speech level X1 detected by the speech level detection unit 103A is From equation (2)
A1 / X1 = 1
And
X1 = A1 = 700
It becomes.

In this case, the updated transmission gain G1 is expressed by the following equation (1):
G1 = G0−0.5 × 20 × log (X1 / 1000) dB
= 0.0-0.5 x 20 x log (700/1000) dB
= 1.2dB
It becomes.

In the utterance at the time of the second speech recognition, the transmission gain before update is G1 = 1.55 dB, and is set in the amplifier 108 before utterance. The detected speech level X2 in this case is
X2 = A2 × 10 ^{G1 / 20}
= 700 × 10 ^{1.55 / 20}
= 837
It becomes.

In this case, the updated transmission gain G2 is calculated from the equation (1):
G2 = G1-0.5 * 20 * log (X2 / 1000) dB
= 1.55-0.5 x 20 x log (837/1000) dB
= 2.32 dB
It becomes.

In the utterance at the time of the third speech recognition, the transmission gain before update is G2 = 2.32 dB and is set in the amplifier 108 before utterance. In this case, the detected speech level detected by the speech level detection unit 103A X3 is calculated from equation (2).
X3 = A3 × 10 ^{G2 / 20}
= 700 × 10 ^{2.32 / 20}
= 914
It becomes.

In this case, the updated transmission gain G3 is calculated from the equation (1):
G3 = G2-0.5 * 20 * log (X3 / 1000) dB
= 2.32-0.5 x 20 x log (914/1000) dB
= 2.71 dB
It becomes.

In the utterance at the time of the fourth speech recognition, the transmission gain before update is G3 = 2.71 dB and is set in the amplifier 108 before utterance. In this case, the speech level X4 detected by the speech level detection unit 103A From the equation (2)
X4 = A4 × 10 ^{G3 / 20}
= 700 × 10 ^{2.71 / 20}
= 956
It becomes.

In this case, the updated transmission gain G4 is expressed by the following equation (1):
G4 = G3−0.5 × 20 × log (X4 / 1000) dB
= 2.71−0.5 × 20 × log (956/1000) dB
= 2.91 dB
It becomes.

In the utterance at the time of the fifth speech recognition, the transmission gain before update is G4 = 2.91 dB and is set in the amplifier 108 before utterance. In this case, the speech level X5 detected by the speech level detection unit 103A From the equation (2)
X5 = A5 × 10 ^{G4 / 20}
= 700 × 10 ^{2.91 / 20}
= 979
It becomes.

In this case, the updated transmission gain G5 is calculated from the equation (1):
G5 = G4−0.5 × 20 × log (X5 / 1000) dB
= 2.91−0.5 × 20 × log (979/1000) dB
= 2.4dB
It becomes.

The final transmission gain in this case is
−20 × log (700/1000) dB
= 3.00 dB
It is.
As described above, as shown in FIG. 4, it is not possible to achieve an appropriate sound level with the second utterance, but it is possible to achieve an approximately appropriate sound level with the fifth utterance. That is, by updating the transmission gain, the optimal transmission gain is obtained.

FIG. 7 is a calculation example of the transmission gain by the transmission gain setting control unit 103B in FIG. 1. The distance from the microphone 109 to the user's mouth (d1 = 3 cm) is small, and the sound level input to the microphone 109 is large. It is a figure explaining an example in case there is no variation and time constant K = 0.5.
In this figure, as an example, the voice level input to the microphone 109 during the first, second, third, fourth, fifth, sixth, etc. speech recognition is A1 = 1300, With A2 = 1300, A3 = 1300, A4 = 1300, A5 = 1300,... And a time constant of 0.5, the transmission gain is calculated as follows.

In the utterance at the time of the first speech recognition, the initial value of the transmission gain is G0 = 0 dB and is set in the amplifier 108 before utterance. In this case, the speech level X1 detected by the speech level detection unit 103A is From equation (2)
A1 / X1 = 1
And
X1 = A1 = 1300
It becomes.

In this case, the updated transmission gain G1 is expressed by the following equation (1):
G1 = G0−0.5 × 20 × log (X1 / 1000) dB
= 0.0-0.5 x 20 x log (1300/1000) dB
= -1.14 dB
It becomes.

In the utterance at the time of the second speech recognition, the transmission gain before update is G1 = −1.14 dB, and is set in the amplifier 108 before utterance. In this case, the detected speech level X2 is expressed by equation (2). Than,
X2 = A2 × 10 ^{G1 / 20}
= 1300 × 10 ^{−1.14 / 20}
= 1140
It becomes.

In this case, the updated transmission gain G2 is calculated from the equation (1):
G2 = G1-0.5 * 20 * log (X2 / 1000) dB
= -1.14-0.5 x 20 x log (1140/1000) dB
= -1.71 dB
It becomes.

In the utterance at the time of the third speech recognition, the transmission gain before update is G2 = -1.71 dB, and is set in the amplifier 108 before utterance. In this case, the detected speech detected by the speech level detection unit 103A Level X3 is calculated from equation (2).
X3 = A3 × 10 ^{G2 / 20}
= 1300 × 10 ^{−1.71 / 20}
= 1068
It becomes.

In this case, the updated transmission gain G3 is calculated from the equation (1):
G3 = G2-0.5 * 20 * log (X3 / 1000) dB
= -1.71-0.5 x 20 x log (1068/1000) dB
= -2.00 dB
It becomes.

In the utterance at the time of the fourth speech recognition, the transmission gain before update is G3 = −2.00 dB, and is set in the amplifier 108 before utterance. In this case, the speech level detected by the speech level detection unit 103A X4 is calculated from equation (2).
X4 = A4 × 10 ^{G3 / 20}
= 1300 × 10 ^{−2.00 / 20}
= 1032
It becomes.

In this case, the updated transmission gain G4 is expressed by the following equation (1):
G4 = G3−0.5 × 20 × log (X4 / 1000) dB
= −2.00−0.5 × 20 × log (1032/1000) dB
= -2.14 dB
It becomes.

In the utterance at the time of the fifth speech recognition, the transmission gain before update is G4 = -2.14 dB, and is set in the amplifier 108 before utterance. In this case, the speech level detected by the speech level detection unit 103A X5 is calculated from equation (2).
X5 = A5 × 10 ^{G4 / 20}
= 1300 × 10 ^{−2.14 / 20}
= 1016
It becomes.

In this case, the updated transmission gain G5 is calculated from the equation (1):
G5 = G4−0.5 × 20 × log (X5 / 1000) dB
= −2.14−0.5 × 20 × log (1016/1000) dB
= -2.21dB
It becomes.

The final transmission gain in this case is
−20 × log (1300/1000) dB
= -2.28 dB
It is.
As described above, as shown in FIG. 4, it is not possible to achieve an appropriate sound level with the second utterance, but it is possible to achieve an approximately appropriate sound level with the fifth utterance. That is, by updating the transmission gain, the optimal transmission gain is obtained.

FIG. 8 is a calculation example of the transmission gain by the transmission gain setting control unit 103B in FIG. 1, and the distance from the microphone 109 to the user's mouth (d2 = 7 cm) is large, and the sound level input to the microphone 109 is small. It is a figure explaining an example in case there is no variation and time constant K = 0.5.
In this figure, as an example, when the first, second, third, fourth, fifth, sixth,... With A2 = 300, A3 = 300, A4 = 300, A5 = 300,... And a time constant of 0.5, the transmission gain is calculated as follows.

In the utterance at the time of the first speech recognition, the initial value of the transmission gain is G0 = 0 dB and is set in the amplifier 108 before utterance. In this case, the speech level X1 detected by the speech level detection unit 103A is From equation (2)
A1 / X1 = 1
And
X1 = A1 = 300
It becomes.

In this case, the updated transmission gain G1 is expressed by the following equation (1):
G1 = G0−0.5 × 20 × log (X1 / 1000) dB
= 0.0-0.5 x 20 x log (300/1000) dB
= 5.23 dB
It becomes.

In the utterance at the time of the second speech recognition, the transmission gain before update is G1 = 5.23 dB, and is set in the amplifier 108 before utterance. In this case, the detected speech level X2 is expressed by the following equation (2):
X2 = A2 × 10 ^{G1 / 20}
= 300 × 10 ^{5.23 / 20}
= 548
It becomes.

In this case, the updated transmission gain G2 is calculated from the equation (1):
G2 = G1-0.5 * 20 * log (X2 / 1000) dB
= 5.23−0.5 × 20 × log (548/1000) dB
= 7.84 dB
It becomes.

In the utterance at the time of the third speech recognition, the transmission gain before update is G2 = 7.84 dB, and is set in the amplifier 108 before utterance. In this case, the detected speech level detected by the speech level detection unit 103A X3 is calculated from equation (2).
X3 = A3 × 10 ^{G2 / 20}
= 300 × 10 ^{7.84 / 20}
= 740
It becomes.

In this case, the updated transmission gain G3 is calculated from the equation (1):
G3 = G2-0.5 * 20 * log (X3 / 1000) dB
= 7.84−0.5 × 20 × log (740/1000) dB
= 8.90 dB
It becomes.

In the utterance at the time of the fourth speech recognition, the transmission gain before update is G3 = 8.90 dB, and is set in the amplifier 108 before utterance. In this case, the speech level X4 detected by the speech level detection unit 103A From the equation (2)
X4 = A4 × 10 ^{G3 / 20}
= 300 × 10 ^{8.90 / 20}
= 836
It becomes.

In this case, the updated transmission gain G4 is expressed by the following equation (1):
G4 = G3−0.5 × 20 × log (X4 / 1000) dB
= 8.90−0.5 × 20 × log (836/1000) dB
= 9.68 dB
It becomes.

In the utterance at the time of the fifth speech recognition, the transmission gain before update is G4 = 9.68 dB, and is set in the amplifier 108 before utterance. In this case, the speech level X5 detected by the speech level detection unit 103A From the equation (2)
X5 = A5 × 10 ^{G4 / 20}
= 300 × 10 ^{9.68 / 20}
= 914
It becomes.

In this case, the updated transmission gain G5 is calculated from the equation (1):
G5 = G4−0.5 × 20 × log (X5 / 1000) dB
= 9.68−0.5 × 20 × log (914/1000) dB
= 10.07 dB
It becomes.

The final transmission gain in this case is
−20 × log (300/1000) dB
= 10.46 dB
It is.
As described above, as shown in FIG. 4, it is not possible to achieve an appropriate sound level with the second utterance, but it is possible to achieve an approximately appropriate sound level with the fifth utterance. That is, by updating the transmission gain, the optimal transmission gain is obtained.

  Next, with the time constant K = 0.5 as it is, there is variation, and the sound level detected by the sound level detection unit 103A is brought closer to the appropriate sound level C as follows.

FIG. 9 is a calculation example of the transmission gain by the transmission gain setting control unit 103B in FIG. 1, and the distance (d1 = 3 cm) from the microphone 109 to the user's mouth is small, and the voice level input to the microphone 109 is small. It is a figure explaining an example in case there is variation and time constant K = 0.5.
In this figure, as an example, when the first, second, third, fourth, fifth, and so on during voice recognition, the voice levels input to the microphone 109 are A1 = 700, A2 = 750, With A3 = 700, A4 = 750, A5 = 700, A6 = 750,... And a time constant of 0.5, the transmission gain is calculated as follows.

In the utterance at the time of the first speech recognition, the initial value of the transmission gain is G0 = 0 dB and is set in the amplifier 108 before utterance. In this case, the speech level X1 detected by the speech level detection unit 103A is From equation (2)
A1 / X1 = 1
And
X1 = A1 = 700
It becomes.

In this case, the updated transmission gain G1 is expressed by the following equation (1):
G1 = G0−0.5 × 20 × log (X1 / 1000) dB
= 0.0-0.5 x 20 x log (700/1000) dB
= 1.55 dB
It becomes.

In the utterance at the time of the second speech recognition, the transmission gain before update is G1 = 1.55 dB, and is set in the amplifier 108 before utterance. The detected speech level X2 in this case is
X2 = A2 × 10 ^{G1 / 20}
= 750 × 10 ^{1.55 / 20}
= 895
It becomes.

In this case, the updated transmission gain G2 is calculated from the equation (1):
G2 = G1-0.5 * 20 * log (X2 / 1000) dB
= 1.55-0.5 x 20 x log (895/1000) dB
= 2.02 dB
It becomes.

In the utterance at the time of the third speech recognition, the transmission gain before update is G2 = 2.02 dB, and is set in the amplifier 108 before utterance. In this case, the detected speech level detected by the speech level detection unit 103A X3 is calculated from equation (2).
X3 = A3 × 10 ^{G2 / 20}
= 700 × 10 ^{2.02 / 20}
= 883
It becomes.

In this case, the updated transmission gain G3 is calculated from the equation (1):
G3 = G2-0.5 * 20 * log (X3 / 1000) dB
= 2.02-0.5 x 20 x log (883/1000) dB
= 2.56dB
It becomes.

In the utterance at the time of the fourth speech recognition, the transmission gain before update is G3 = 2.56 dB, and is set in the amplifier 108 before utterance. In this case, the speech level X4 detected by the speech level detection unit 103A From the equation (2)
X4 = A4 × 10 ^{G3 / 20}
= 750 × 10 ^{2.56 / 20}
= 1007
It becomes.

In this case, the updated transmission gain G4 is expressed by the following equation (1):
G4 = G3−0.5 × 20 × log (X4 / 1000) dB
= 2.56-0.5 × 20 × log (1007/1000) dB
= 2.53 dB
It becomes.

In the utterance at the time of the fifth speech recognition, the transmission gain before update is G4 = 2.53 dB, and is set in the amplifier 108 before utterance. In this case, the speech level X5 detected by the speech level detection unit 103A From the equation (2)
X5 = A5 × 10 ^{G4 / 20}
= 700 × 10 ^{2.53 / 20}
= 937
It becomes.

In this case, the updated transmission gain G5 is calculated from the equation (1):
G5 = G4−0.5 × 20 × log (X5 / 1000) dB
= 2.53-0.5 × 20 × log (937/1000) dB
= 2.81 dB
It becomes.

The explanation of the process from the 6th to the 8th in the figure is omitted, and the audio levels “954” and “1046” detected by the audio level detection unit 103A in the 9th and 10th utterances are transmitted. The talk gains “2.89” and “2.69” are repeated in subsequent utterances.
As compared with the case where the time constant K = 1.0 in FIG. 5 is set in this way, when the time constant K = 0.5, the sound level detected by the sound level detection unit 103A is the appropriate sound level C (= 1000). ) Requires a greater number of utterances, but approaches the appropriate voice level and improves the recognition rate of voice recognition. That is, by updating the transmission gain, the optimum transmission gain is approached.

FIG. 10 is a calculation example of the transmission gain by the transmission gain setting control unit 103B in FIG. 1. The distance from the microphone 109 to the user's mouth (d2 = 7 cm) is large, and the sound level input to the microphone 109 is small. It is a figure explaining the example when there is no variation, time constant K = 0.5, and noise suddenly enters in a state where the sound level detected by the sound level detection unit 103A is close to the appropriate sound level.
In this figure, as an example, the voice level input to the microphone 109 when the first, second, third, fourth, fifth, sixth,. , A2 = 500, A3 = 300, A4 = 300, A5 = 300, A6 = 300, A7 = 300,..., And the time constant is 0.5, the transmission gain is calculated as follows.

That is, the voice level input to the microphone 109 at the first utterance when voice recognition is resumed is A1 = 300, and the voice level detected by the voice level detection unit 103A is almost the appropriate voice level C = 1000. Suppose that noise is input only in the second time.
In the utterance at the time of the first speech recognition, the transmission gain before update is G0 = 10.46 dB and is set in the amplifier 108 before utterance. In this case, the speech level X1 detected by the speech level detection unit 103A From the equation (2)
X1 = A1 × 10 ^{G0 / 20}
= 700 × 10 ^{10.46 / 20}
= 1000
It becomes.

In this case, the updated transmission gain G1 is expressed by the following equation (1):
G1 = G0−0.5 × 20 × log (X1 / 1000) dB
= 10.46-0.5 x 20 x log (1000/1000) dB
= 10.46 dB
It becomes.

In the utterance at the time of the second speech recognition, the transmission gain before the update is G1 = 10.46 dB, and is set in the amplifier 108 before the utterance. The detected speech level X2 in this case is
X2 = A2 × 10 ^{G1 / 20}
= 500 × 10 ^{2.71 / 20}
= 1667
It becomes.

In this case, the updated transmission gain G2 is calculated from the equation (1):
G2 = G1-0.5 * 20 * log (X2 / 1000) dB
= 10.46-0.5 x 20 x log (1667/1000) dB
= 8.24 dB
It becomes.

In the utterance at the time of the third speech recognition, the transmission gain before update is G2 = 8.24 dB, and is set in the amplifier 108 before utterance. In this case, the detected speech level detected by the speech level detection unit 103A X3 is calculated from equation (2).
X3 = A3 × 10 ^{G2 / 20}
= 300 × 10 ^{8.24 / 20}
= 775
It becomes.

In this case, the updated transmission gain G3 is calculated from the equation (1):
G3 = G2-0.5 * 20 * log (X3 / 1000) dB
= 8.24-0.5 × 20 × log (775/1000) dB
= 9.35 dB
It becomes.

In the utterance at the time of the fourth speech recognition, the transmission gain before update is G3 = 9.35 dB and is set in the amplifier 108 before utterance. In this case, the speech level X4 detected by the speech level detection unit 103A From the equation (2)
X4 = A4 × 10 ^{G3 / 20}
= 300 × 10 ^{9.35 / 20}
= 880
It becomes.

In this case, the updated transmission gain G4 is expressed by the following equation (1):
G4 = G3−0.5 × 20 × log (X4 / 1000) dB
= 9.35-0.5 x 20 x log (880/1000) dB
= 9.91 dB
It becomes.

In the utterance at the time of the fifth speech recognition, the transmission gain before update is G4 = 9.91 dB, and is set in the amplifier 108 before utterance. In this case, the speech level X5 detected by the speech level detector 103A From the equation (2)
X5 = A5 × 10 ^{G4 / 20}
= 300 × 10 ^{9.91 / 20}
= 939
It becomes.

In this case, the updated transmission gain G5 is calculated from the equation (1):
G5 = G4−0.5 × 20 × log (X5 / 1000) dB
= 9.91−0.5 × 20 × log (939/1000) dB
= 10.18 dB
It becomes.

In the utterance at the time of the sixth speech recognition, the transmission gain before update is G5 = 10.18 dB, and is set in the amplifier 108 before utterance. In this case, the speech level X6 detected by the speech level detection unit 103A From the equation (2)
X6 = A6 × 10 ^{G5 / 20}
= 300 × 10 ^{10.18 / 20}
= 969
It becomes.

In this case, the updated transmission gain G6 is as follows from the equation (1):
G6 = G5−0.5 × 20 × log (X6 / 1000) dB
= 10.18-0.5 x 20 x log (969/1000) dB
= 10.32 dB
It becomes.
Thus, even if the transmission gain changes due to noise, the transmission gain is updated to return to the original optimum transmission gain.

FIG. 11 is an example of calculation of the transmission gain by the transmission gain setting control unit 103B in FIG. 1, in which the distance from the microphone 109 to the user's mouth (d2 = 7 cm) is large , the sound level input to the microphone 109 is small , and variation occurs. There is no time constant, and the time constant K = 0.5, and the distance from the microphone 109 to the user's mouth (d1 = 3 cm) is suddenly small when the sound level detected by the sound level detection unit 103A is close to the appropriate sound level. It is a figure explaining the example when the audio | voice level input into the microphone 109 becomes large .

In this figure, as an example, the voice level input to the microphone 109 when the first, second, third, fourth, fifth, sixth, seventh,... Where A1 = 300, A2 = 700, A3 = 700, A4 = 700, A5 = 700, A6 = 700, A7 = 700,..., And the time constant is 0.5, the transmission gain is calculated as follows: To do.
That is, the voice level input to the microphone 109 at the first utterance when voice recognition is resumed is A1 = 300, and the voice level detected by the voice level detection unit 103A is almost the appropriate voice level C = 1000. Assume that the audio level after the second time changes to 700.

In the utterance at the time of the first speech recognition, the transmission gain before update is G0 = 10.46 dB and is set in the amplifier 108 before utterance. In this case, the speech level X1 detected by the speech level detection unit 103A From the equation (2)
X1 = A1 × 10 ^{G0 / 20}
= 300 × 10 ^{10.46 / 20}
= 1000
It becomes.

In this case, the updated transmission gain G1 is expressed by the following equation (1):
G1 = G0−0.5 × 20 × log (X1 / 1000) dB
= 10.46-0.5 x 20 x log (1000/1000) dB
= 10.46 dB
It becomes.

In the utterance at the time of the second speech recognition, the transmission gain before the update is G1 = 10.46 dB, and is set in the amplifier 108 before the utterance. The detected speech level X2 in this case is
X2 = A2 × 10 ^{G1 / 20}
= 700 × 10 ^{10.46 / 20}
= 2331
It becomes.

In this case, the updated transmission gain G2 is calculated from the equation (1):
G2 = G1-0.5 * 20 * log (X2 / 1000) dB
= 10.46-0.5 x 20 x log (2331/1000) dB
= 6.78 dB
It becomes.

In the utterance at the time of the third speech recognition, the transmission gain before update is G2 = 6.78 dB, and is set in the amplifier 108 before utterance. In this case, the detected speech level detected by the speech level detection unit 103A X3 is calculated from equation (2).
X3 = A3 × 10 ^{G2 / 20}
= 700 × 10 ^{6.78 / 20}
= 1528
It becomes.

In this case, the updated transmission gain G3 is calculated from the equation (1):
G3 = G2-0.5 * 20 * log (X3 / 1000) dB
= 6.78−0.5 × 20 × log (1528/1000) dB
= 4.94 dB
It becomes.

In the utterance at the time of the fourth speech recognition, the transmission gain before update is G3 = 4.94 dB, and is set in the amplifier 108 before utterance. In this case, the speech level X4 detected by the speech level detection unit 103A From the equation (2)
X4 = A4 × 10 ^{G3 / 20}
= 700 × 10 ^{4.94 / 20}
= 1236
It becomes.

In this case, the updated transmission gain G4 is expressed by the following equation (1):
G4 = G3−0.5 × 20 × log (X4 / 1000) dB
= 4.94−0.5 × 20 × log (1236/1000) dB
= 4.02 dB
It becomes.

In the utterance at the time of the fifth speech recognition, the transmission gain before update is G4 = 4.02 dB, and is set in the amplifier 108 before utterance. In this case, the speech level X5 detected by the speech level detection unit 103A From the equation (2)
X5 = A5 × 10 ^{G4 / 20}
= 700 × 10 ^{4.02 / 20}
= 1112
It becomes.

In this case, the updated transmission gain G5 is calculated from the equation (1):
G5 = G4−0.5 × 20 × log (X5 / 1000) dB
= 4.02-0.5 x 20 x log (1112/1000) dB
= 3.56dB
It becomes.

In the utterance at the time of the sixth speech recognition, the transmission gain before update is G5 = 3.56 dB, and is set in the amplifier 108 before utterance. In this case, the speech level X6 detected by the speech level detection unit 103A From the equation (2)
X6 = A6 × 10 ^{G5 / 20}
= 700 × 10 ^{3.56 / 20}
= 1055
It becomes.

In this case, the updated transmission gain G6 is as follows from the equation (1):
G6 = G5−0.5 × 20 × log (X6 / 1000) dB
= 3.56-0.5 × 20 × log (1055/1000) dB
= 3.33dB
It becomes.

In the utterance at the time of the seventh speech recognition, the transmission gain before update is G6 = 3.33 dB, and is set in the amplifier 108 before utterance. In this case, the speech level X7 detected by the speech level detection unit 103A From the equation (2)
X7 = A7 × 10 ^{G6 / 20}
= 700 × 10 ^{3.33 / 20}
= 1027
It becomes.

In this case, the updated transmission gain G7 is calculated from the equation (1):
G7 = G6−0.5 × 20 × log (X7 / 1000) dB
= 3.33-0.5 x 20 x log (1027/1000) dB
= 3.21 dB
It becomes.

  As described above, the distance from the microphone 109 to the user's mouth (d2 = 7 cm) is large, the sound level input to the microphone 109 is small, and the sound level detected by the sound level detection unit 103A is close to the appropriate sound level. Even when the distance from the microphone 109 to the user's mouth (d1 = 3 cm) is suddenly small and the voice level input to the microphone 109 becomes large, the transmission gain is updated, so that Optimal transmission gain.

FIG. 12 is a calculation example of the transmission gain by the transmission gain setting control unit 103B in FIG. 1. The distance (d1 = 3 cm) from the microphone 109 to the user's mouth is small , and the sound level input to the microphone 109 is large . There is no variation, the time constant K = 0.5, and the distance from the microphone 109 to the user's mouth (d2 = 7 cm) suddenly becomes a state where the sound level detected by the sound level detection unit 103A is close to the appropriate sound level. large, it is a diagram illustrating an example of a case where the sound level was reduced to be inputted to the microphone 109.

In this figure, as an example, the voice level input to the microphone 109 when the first, second, third, fourth, fifth, sixth, seventh,... Where A1 = 700, A2 = 300, A3 = 300, A4 = 300, A5 = 300, A6 = 300, A7 = 300,..., And the time constant is 0.5, the transmission gain is calculated as follows: To do.
That is, the voice level input to the microphone 109 at the time of the first utterance when resuming voice recognition is A1 = 700, and the voice level detected by the voice level detector 103A is substantially the appropriate voice level C = 1000. Suppose that the audio level after the second time changes to 300.

In the utterance at the time of the first speech recognition, the transmission gain before update is G0 = 3.00 dB and is set in the amplifier 108 before utterance. In this case, the speech level X1 detected by the speech level detection unit 103A From the equation (2)
X1 = A1 × 10 ^{G0 / 20}
= 700 × 10 ^{3.00 / 20}
= 1000
It becomes.

In this case, the updated transmission gain G1 is expressed by the following equation (1):
G1 = G0−0.5 × 20 × log (X1 / 1000) dB
= 3.00-0.5 x 20 x log (1000/1000) dB
= 3.00 dB
It becomes.

In the utterance at the time of the second speech recognition, the transmission gain before the update is G1 = 3.00 dB, and is set in the amplifier 108 before the utterance. In this case, the detected speech level X2 is obtained from the equation (2):
X2 = A2 × 10 ^{G1 / 20}
= 300 × 10 ^{3.00 / 20}
= 424
It becomes.

In this case, the updated transmission gain G2 is calculated from the equation (1):
G2 = G1-0.5 * 20 * log (X2 / 1000) dB
= 3.00-0.5 x 20 x log (424/1000) dB
= 6.73 dB
It becomes.

In the utterance at the time of the third speech recognition, the transmission gain before update is G2 = 6.73 dB, and is set in the amplifier 108 before utterance. In this case, the detected speech level detected by the speech level detection unit 103A X3 is calculated from equation (2).
X3 = A3 × 10 ^{G2 / 20}
= 300 × 10 ^{6.73 / 20}
= 651
It becomes.

In this case, the updated transmission gain G3 is calculated from the equation (1):
G3 = G2-0.5 * 20 * log (X3 / 1000) dB
= 6.73−0.5 × 20 × log (651/1000) dB
= 8.60 dB
It becomes.

In the utterance at the time of the fourth speech recognition, the transmission gain before update is G3 = 8.60 dB and is set in the amplifier 108 before utterance. In this case, the speech level X4 detected by the speech level detection unit 103A From the equation (2)
X4 = A4 × 10 ^{G3 / 20}
= 300 × 10 ^{8.60 / 20}
= 807
It becomes.

In this case, the updated transmission gain G4 is expressed by the following equation (1):
G4 = G3−0.5 × 20 × log (X4 / 1000) dB
= 8.60−0.5 × 20 × log (807/1000) dB
= 9.53 dB
It becomes.

In the utterance at the time of the fifth speech recognition, the transmission gain before update is G4 = 9.53 dB, and is set in the amplifier 108 before utterance. In this case, the speech level X5 detected by the speech level detection unit 103A From the equation (2)
X5 = A5 × 10 ^{G4 / 20}
= 300 × 10 ^{9.53 / 20}
= 899
It becomes.

In this case, the updated transmission gain G5 is calculated from the equation (1):
G5 = G4−0.5 × 20 × log (X5 / 1000) dB
= 9.53−0.5 × 20 × log (899/1000) dB
= 9.99 dB
It becomes.

In the utterance at the time of the sixth speech recognition, the transmission gain before update is G5 = 9.99 dB, and is set in the amplifier 108 before utterance. In this case, the speech level X6 detected by the speech level detection unit 103A From the equation (2)
X6 = A6 × 10 ^{G5 / 20}
= 300 × 10 ^{9.99 / 20}
= 948
It becomes.

In this case, the updated transmission gain G6 is as follows from the equation (1):
G6 = G5−0.5 × 20 × log (X6 / 1000) dB
= 9.99−0.5 × 20 × log (948/1000) dB
= 10.22 dB
It becomes.

In the utterance at the time of the seventh speech recognition, the transmission gain before update is G6 = 10.22 dB, and is set in the amplifier 108 before utterance. In this case, the speech level X7 detected by the speech level detection unit 103A From the equation (2)
X7 = A7 × 10 ^{G6 / 20}
= 300 × 10 ^{10.22 / 20}
= 973
It becomes.

In this case, the updated transmission gain G7 is calculated from the equation (1):
G7 = G6−0.5 × 20 × log (X7 / 1000) dB
= 10.22-0.5 × 20 × log (973/1000) dB
= 10.34 dB
It becomes.

  As described above, the distance from the microphone 109 to the user's mouth (d1 = 3 cm) is small, the sound level input to the microphone 109 is large, and the sound level detected by the sound level detection unit 103A is close to the appropriate sound level. Even when the distance from the microphone 109 to the user's mouth (d2 = 7 cm) suddenly increases and the sound level input to the microphone 109 decreases, the transmission gain is updated, so that Optimal transmission gain.

FIG. 13 is a flowchart for explaining another series of operation examples of the transmission gain update processing of the transmission gain setting control unit 103B in FIG.
As shown in this figure, in step 211, the control unit 103 detects that the voice recognition start key of the operation unit 104 is pressed and detects that voice recognition is activated.
In step 212, the transmission gain setting control unit 103B reads the initial value (G0) of the transmission gain from the transmission gain information storage unit 106A and sets it in the amplifier 108 after detecting the activation of the control unit 103.

In step 213, after the transmission gain setting control unit 103 B sets the transmission gain in the amplifier 108, the control unit 103 adjusts the voice signal input from the microphone 109 to a voice level suitable for voice recognition by the amplifier 108. For the input voice, the voice recognition unit 111 is activated to perform voice recognition.
In step 214, the transmission gain setting control unit 103B waits for confirmation of the recognition result from the speech recognition unit 111.

In step 215, the transmission gain setting control unit 103B acquires the recognition result and the voice level information detected by the voice level detection unit 103A after the recognition result is confirmed. The recognition result is displayed on the display unit 105, and the speaker 110 is made to output the sound.
In step 216, the transmission gain setting control unit 103B starts a timer and measures the time (TIME) until the next voice recognition activation detection.

In step 217, the transmission gain setting control unit 103B compares the measured time TIME with the predetermined time Th,
TIME> Th
If is established, the process is terminated. That is, when the voice recognition activation is not detected again even after the predetermined time Th has elapsed, the process is terminated as the voice recognition is completed.

In step 218, the control unit 103 detects whether or not the voice recognition start key of the operation unit 104 is pressed and detects whether or not voice recognition is activated. If activation is not detected, the process returns to step 217.
In step 219, the transmission gain setting control unit 103B detects the activation of the control unit 103, reads the initial value (G0) of the transmission gain from the transmission gain information storage unit 106A, sets it in the amplifier 108, and acquires the acquired voice level. Based on the information, the initial value of the transmission gain is updated and stored in the transmission gain information storage unit 106A, or the transmission gain (Gn-1) before update is read from the transmission gain information storage unit 106A and set in the amplifier 108. The transmission gain is updated based on the acquired voice level information, the updated transmission gain (Gn) is stored in the transmission gain information storage unit 106A, and the updated transmission gain at the next utterance is stored in the transmission gain information. The read amplifier 108 is set from the unit 106A.

In step 220, after the transmission gain setting control unit 103B sets the transmission gain in the amplifier 108, the control unit 103 adjusts the audio signal input from the microphone 109 to an audio level suitable for speech recognition by the amplifier 108. For the input voice, the voice recognition unit 111 is activated to perform voice recognition.
In step 221, the transmission gain setting control unit 103B waits for a speech recognition process until the speech recognition unit 111 determines a speech recognition result.

In step 222, the transmission gain setting control unit 103B acquires the recognition result and the voice level information detected by the voice level detection unit 103A after the recognition result is confirmed. The recognition result is displayed on the display unit 105 and is output to the speaker 110 by voice, and the process returns to step 216.
In this way, the value held in the transmission gain information storage unit 106A is always used as the initial value of the transmission gain when the voice is not continuously activated, and the transmission gain is updated when the voice recognition is continuously activated. Perform gain optimization. That is, when an erroneous recognition occurs in the previous speech recognition, the speech recognition is restarted, and the transmission gain is updated at the restart.

FIG. 14 is a flowchart for explaining still another operation example of a series of transmission gain update processing of the transmission gain setting control unit 103B in FIG.
As shown in the figure, in step 231, the control unit 103 detects that the voice recognition start key of the operation unit 104 is pressed and detects that voice recognition is activated.
In step 232, the transmission gain setting control unit 103B detects the activation of the control unit 103, reads the transmission gain (Gn-1) before update from the transmission gain information storage unit 106A, and sets it in the amplifier 108. When the transmission gain (Gn-1) before update is not held in the transmission gain information storage unit 106A, the initial value (G0) of the transmission gain is set in the amplifier 108.

In step 233, the audio level detection unit 103A detects the audio level Xn,
900 ≦ Xn ≦ 1100
(Proper audio level C = 1000)
Whether or not the inequality is satisfied.
In step 234, when this inequality is satisfied, the time constant K is set to 0.5 and the process proceeds to step 236.

In step 235, if this inequality is not satisfied, the time constant K is set to 1.0.
In step 236, the control unit 103 sets the transmission gain by the transmission gain setting control unit 103B in the amplifier 108, and then adjusts the voice signal input to the microphone 109 to a voice level suitable for voice recognition by the amplifier 108. In response to this, the voice recognition unit 111 is activated to perform voice recognition.

In step 237, the transmission gain setting control unit 103B waits for confirmation of the recognition result from the speech recognition unit 111.
In step 238, the transmission gain setting control unit 103B acquires the recognition result and the voice level information detected by the voice level detection unit 103A after the recognition result is confirmed. The recognition result is displayed on the display unit 105, and the speaker 110 is made to output the sound.

In step 239, transmission gain update processing (see equation (1)) is performed. In this case, the time constant K determined in step 234 and step 235 is set.
In step 240, the transmission gain setting control unit 103B stores the updated transmission gain (Gn) value in the transmission gain information storage unit 106A, and ends the process.
In this way, when the audio level Xn detected by the audio level detection unit 103A is far from the appropriate audio level C, the time constant K is increased, and the audio level Xn is close to the appropriate audio level C. In this case, the time constant K is reduced, the number of utterances until the appropriate voice level is reached, and the optimum transmission gain can be acquired.

FIG. 15 is a block diagram showing a schematic configuration of a portable information terminal device according to a modification of FIG. As shown in this figure, compared to FIG. 1, a test unit 103C is provided in the control unit 103, and the test unit 103C can adjust a time constant by operating a key in a test mode on the operation unit 104. The test “test” speech recognition is performed a plurality of times, the speech recognition unit 111 processes the speech recognition, the transmission gain setting control unit 103B determines the optimum value of the transmission gain in advance, and the transmission gain information The storage unit 106A may store the transmission gain as an initial value.
Thereby, the optimum transmission gain at the time of voice recognition can be easily set.

  In the above description, the portable information terminal device has been described. However, the portable information terminal device is not limited to a mobile phone, a PHS (simple phone), a pager, an electronic notebook, or any other mobile device including a personal computer. The invention can be used.

It is a block diagram which shows schematic structure of the portable information terminal device which concerns on this invention. It is a figure explaining the example of various information hold | maintained at the transmission gain information storage part 106A in FIG. 1, and optimizing a transmission gain at the time of speech recognition. 3 is a flowchart for explaining a series of operation examples of transmission gain update processing of a transmission gain setting control unit 103B in FIG. In the example of calculation of the transmission gain by the transmission gain setting control unit 103B in FIG. 1, the distance from the microphone 109 to the user's mouth (d1 = 3 cm) is small, the voice level input to the microphone 109 is small, and there is no variation. It is a figure explaining the example in case the time constant K = 1.0. In the example of calculation of the transmission gain by the transmission gain setting control unit 103B in FIG. 1, the distance from the microphone 109 to the user's mouth (d1 = 3 cm) is small, the voice level input to the microphone 109 is small, and there are variations. It is a figure explaining an example in case of time constant K = 1.0. In the example of calculation of the transmission gain by the transmission gain setting control unit 103B in FIG. 1, the distance from the microphone 109 to the user's mouth (d1 = 3 cm) is small, the voice level input to the microphone 109 is small, and there is no variation. It is a figure explaining the example in case the time constant K = 0.5. In the example of calculation of the transmission gain by the transmission gain setting control unit 103B in FIG. It is a figure explaining the example in case the time constant K = 0.5. In the example of calculation of the transmission gain by the transmission gain setting control unit 103B in FIG. 1, the distance from the microphone 109 to the user's mouth (d2 = 7 cm) is large, the voice level input to the microphone 109 is small, and there is no variation. It is a figure explaining the example in case the time constant K = 0.5. In the transmission gain calculation example by the transmission gain setting control unit 103B in FIG. 1, the distance from the microphone 109 to the user's mouth (d1 = 3 cm) is small, the voice level input to the microphone 109 is small, and there is variation. It is a figure explaining the example in case the time constant K = 0.5. In the example of calculation of the transmission gain by the transmission gain setting control unit 103B in FIG. 1, the distance from the microphone 109 to the user's mouth (d2 = 7 cm) is large, the voice level input to the microphone 109 is small, and there is no variation. FIG. 10 is a diagram for explaining an example when noise suddenly enters in a state where the time constant K = 0.5 and the sound level detected by the sound level detection unit 103A is close to the appropriate sound level. In the calculation example of the transmission gain by the transmission gain setting control unit 103B in FIG. 1, (d2 = 7 cm) from the microphone 109 to the user's mouth is large , the sound level input to the microphone 109 is small , and there is no variation. The time constant K = 0.5, the distance from the microphone 109 to the user's mouth (d1 = 3 cm) is suddenly small with the sound level detected by the sound level detection unit 103A close to the appropriate sound level, and the microphone 109 It is a figure explaining the example when the audio | voice level input into is increased . In the example of the transmission gain calculation by the transmission gain setting control unit 103B in FIG. 1, the distance from the microphone 109 to the user's mouth (d1 = 3 cm) is small , the sound level input to the microphone 109 is large , and there is no variation. The time constant K = 0.5, and the distance from the microphone 109 to the user's mouth (d2 = 7 cm) suddenly increases with the sound level detected by the sound level detection unit 103A close to the appropriate sound level. It is a figure explaining the example when the audio | voice level input into 109 becomes small . 6 is a flowchart for explaining another series of operation examples of transmission gain update processing of the transmission gain setting control unit 103B in FIG. 10 is a flowchart for explaining still another example of a series of operations for updating the transmission gain of the transmission gain setting control unit 103B in FIG. It is a block diagram which shows schematic structure of the portable information terminal device which concerns on the modification in FIG. It is a figure explaining the usage example of the speech recognition in the mobile telephone used as the premise of this invention.

Explanation of symbols

DESCRIPTION OF SYMBOLS 100 ... Portable information terminal device 101 ... Antenna 102 ... Radio | wireless part 103 ... Control part 103A ... Voice level detection part 103B ... Transmission gain setting control part 103C ... Test part 104 ... Operation part 105 ... Display part 106 ... Memory 106A ... Transmission Talk gain information storage unit 107 ... A / D / D / A converter 108 ... Amplifier
109 ... Microphone 110 ... Speaker 111 ... Voice recognition unit

Claims (9)

In a speech recognition device that recognizes speech input to the microphone of the transmitter,
An amplifier that amplifies the audio signal output from the microphone of the transmission unit by a transmission gain;
A sound level detector for detecting a sound level amplified by the amplifier;
A transmission gain information storage unit that stores an initial value of a transmission gain, a transmission gain, an appropriate voice level, and a transmission gain update time constant for updating the transmission gain ;
Read the transmission gain, appropriate audio level, and transmission gain update time constant from the transmission gain information storage unit, set the transmission gain to the amplifier, and determine the audio level detected by the audio level detection unit A value obtained by multiplying the gain to be set to an appropriate voice level by the transmission gain update time constant is added to the transmission gain to update the transmission gain, and the updated transmission gain is stored in the transmission gain information storage. Transmission gain setting control unit to be stored in the unit,
A voice recognition unit that performs voice recognition by inputting the voice signal amplified by the amplifier;
The time constant for transmission gain update by key operation can be adjusted, the voice recognition unit processes voice recognition for a plurality of test utterances, and the transmission gain setting control unit has an optimum value of the transmission gain. And a test unit that stores the transmission gain information storage unit as an initial value of the transmission gain. 2. The transmission gain setting control unit reads an initial value of the transmission gain from the transmission gain information storage unit and sets it in the amplifier at the first utterance during speech recognition. The speech recognition apparatus described in 1. When the updated transmission gain is stored in the transmission gain information storage unit, the transmission gain setting control unit stores the transmission gain information at the first utterance when speech recognition is resumed. The speech recognition apparatus according to claim 1, wherein the transmission gain updated from a unit is read and set in the amplifier. The transmission gain setting control unit reads the initial value of the transmission gain from the transmission gain information storage unit at the time of the first utterance at the time of voice recognition, sets the value to the amplifier, and performs voice recognition on the voice recognition unit. If the activation of voice recognition is detected within a predetermined time after the voice recognition is confirmed, the initial value of the transmission gain is read from the transmission gain information storage unit and set in the amplifier, and the acquired voice level information The initial value of the transmission gain is updated based on the transmission gain information and stored in the transmission gain information storage unit, or the transmission gain before update is read from the transmission gain information storage unit and set in the amplifier, and the acquired voice level The transmission gain is updated based on the information, stored in the transmission gain information storage unit, the transmission gain updated at the next utterance is read from the transmission gain information storage unit, set in the amplifier, Characterized in that to terminate the voice recognition process if the activation of the voice recognition is not detected in time, the speech recognition apparatus according to claim 1. The transmission gain setting control unit determines whether the voice level detected by the transmission gain information storage unit is within a certain range centered on the appropriate voice level or outside the certain range, characterized by less than time constant for transmission gain update when there a time constant for transmission gain update when located within a predetermined range out of a predetermined range, voice recognition device according to claim 1 . The transmission gain Gn updated by the transmission gain setting control unit is expressed by the following equation: Gn = Gn-1−K × 20 × log (Xn / C) dB
(K: Time constant for transmission gain update for transmission gain update (0.0
<K ≦ 1.0),
n: Number of voice recognition (= 1, 2, 3 ...),
C: Appropriate audio level,
Xn: voice level detected by the voice level detector
The speech recognition apparatus according to claim 1, wherein In a speech recognition method for recognizing speech input to a microphone of a transmitter,
Storing a proper voice level and transmission gain;
Amplifying the audio signal output from the microphone of the transmission unit by the transmission gain;
Detecting the amplified audio level;
A value obtained by multiplying the detected speech level to the appropriate speech level by a transmission gain update time constant is added to the transmission gain to update the transmission gain and store the updated transmission gain. And a process of
Inputting the amplified audio signal and performing speech recognition;
A process for adjusting the time constant for updating the transmission gain by key operation, processing speech recognition for a plurality of test utterances, obtaining the optimum value of the transmission gain in advance, and setting it as the initial value of the transmission gain A speech recognition method comprising: In a portable information terminal device using a speech recognition method,
A portable information terminal device using a voice recognition method, comprising a voice recognition function based on the method according to claim 7 in addition to the portable information terminal function of the portable information terminal device. The speech recognition method according to claim 8 , wherein the recognition result recognized by speech is caused to ring on a speaker of an earpiece of the portable information terminal device and displayed on a display unit of the portable information terminal device. The portable information terminal device used.

Images