WO2016067644A1

WO2016067644A1 - Speech adjustment device

Info

Publication number: WO2016067644A1
Application number: PCT/JP2015/055093
Authority: WO
Inventors: 中村　圭介
Original assignee: シャープ株式会社
Priority date: 2014-10-29
Filing date: 2015-02-23
Publication date: 2016-05-06
Also published as: JP2016085420A; JP5863928B1

Abstract

Provided is a speech adjustment device capable of improving a speech recognition rate by suitably adjusting the input gain and the output gain of a speech signal responsive to usage environment. This speech adjustment device (20) is provided with a speech adjustment unit (40) for adjusting a speech signal by comparing a soundless period or a sound period with a predetermined time threshold, the soundless period being the continuation of a soundless state where the intensity of the speech signal is less than a speech intensity threshold, and the sound period being the continuation of a sound state where the intensity of the speech signal is greater than the speech intensity threshold.

Description

Audio adjustment device

The present invention relates to a sound adjustment device.

In recent years, various devices such as robots and smartphones are equipped with a voice recognition function, and the number of devices that execute instructions by voice dialogue with users is increasing. In such a voice interactive device, in order to improve the voice recognition rate, it is desirable that the sensitivity (input gain) of the microphone is automatically and sequentially adjusted according to the surrounding noise conditions.

For example, in voice recognition, if the microphone sensitivity is too low, the necessary voice signal cannot be obtained sufficiently and voice recognition cannot be performed correctly. In speech recognition, even if the sensitivity of the microphone is too high, distortion of the speech signal and ambient noise are picked up and speech recognition cannot be performed correctly.

Therefore, an audio adjustment device that adjusts the sensitivity of the microphone and maintains it in an appropriate state has been put into practical use.

For example, the speech recognition apparatus disclosed in Patent Document 1 detects an average speech speech included in an input speech signal in a speech speech interval after detecting a time interval in which the user speech speech is included in the input speech data as a speech speech interval. The level and the level of noise included in the input voice signal in the time interval other than the utterance voice interval are calculated, the input voice signal level is estimated from the noise level and the average utterance voice level, and the input voice The gain of the input amplifier is set so that the level obtained by amplifying the signal level with the input amplifier becomes a level suitable for speech recognition.

Japanese Patent Gazette “Patent No. 5457293 (Registered on January 17, 2014)”

However, in the speech recognition apparatus of Patent Document 1, it is necessary to accurately discriminate between a section in which the user's utterance voice is included in the input voice data and other sections, but such utterance voice is generally discriminated. It is very difficult.

The voice recognition device of Patent Document 1 includes a talk switch, and determines that the period from when the user presses the talk switch down to the end of the utterance voice section is the section including the user's utterance voice. Depressing the talk switch for each input has the problem of complicating user operations.

Further, in a general voice recognition apparatus that does not include a talk switch, it is necessary to determine the start time of the speech voice section by another means. Furthermore, in general, voice input data includes a user's speech signal and a noise signal, and it is difficult to determine when only the user's speech signal has ended.

In the speech recognition apparatus disclosed in Patent Document 1, when the input speech data is not accurately discriminated between the section in which the user's speech voice is included and the other sections, the average speech speech included in the input speech signal in the speech speech section And the noise level included in the input voice signal in a time period other than the speech voice period may be erroneously calculated, and the input voice signal level cannot be adjusted appropriately according to the usage environment. There was a problem.

The present invention has been made in view of the above problems, and an object of the present invention is to appropriately adjust the input gain and output gain of a sound signal in accordance with the use environment and improve the sound recognition rate. Is to provide.

The voice adjustment device of the present invention includes a microphone signal input unit that converts an electrical signal input from a microphone into a voice signal, a voice signal strength measurement unit that measures the strength of the voice signal and compares it with a preset voice strength threshold. , A silent time measurement unit that measures a silent period in which a silent state where the voice signal is smaller than a voice intensity threshold continues, and a voiced time that measures a voiced time in which a voice signal whose voice signal is greater than the voice intensity threshold continues A measurement unit and an audio adjustment unit that adjusts the audio signal by comparing the silent time or the sound time with a preset time threshold value are provided.

In the audio adjustment device of the present invention, the audio adjustment unit includes a microphone sensitivity adjustment unit for adjusting the input gain of the microphone signal input unit, and the input gain is set when the silence time is longer than a preset silence time threshold. The input gain is lowered when the sound duration is longer than a preset sound duration threshold.

In the audio adjustment device of the present invention, the audio adjustment unit includes a speaker signal output unit that outputs an audio signal to the speaker, and a speaker volume adjustment unit for adjusting the output gain of the speaker signal output unit, and is a silent time The output gain is lowered when the time is longer than the preset silent time threshold, and the output gain is increased when the voiced time is longer than the preset voiced time threshold.

In the audio adjustment device of the present invention, the audio adjustment unit includes a microphone sensitivity adjustment unit for adjusting the input gain of the microphone signal input unit, a speaker signal output unit that outputs an audio signal to the speaker, and a speaker signal output unit. A speaker volume adjustment unit for adjusting the output gain of the sound, and when the silence time is longer than a preset silent time threshold, the input gain is increased and the output gain is lowered to set the voice time with a preset voice time. When it is longer than the threshold value, the input gain is decreased and the output gain is increased.

According to the present invention, by adjusting the input gain and output gain based on the voiced time and silent time, the input gain and output gain of the audio signal are appropriately adjusted according to the usage environment, and the voice recognition rate is increased. An audio adjustment device that can be improved can be provided.

1 is a basic configuration diagram of an audio adjustment device of the present invention. It is a block diagram of the audio | voice adjustment apparatus of Example 1. FIG. It is a principle figure which shows the adjustment method of the microphone sensitivity using a silence time. It is a principle figure which shows the adjustment method of the microphone sensitivity using sound time. FIG. 6 is a flowchart illustrating a microphone sensitivity adjustment step according to the first embodiment. It is a block diagram of the audio | voice adjustment apparatus of Example 2. FIG. It is a flowchart which shows the adjustment step of the speaker volume of Example 2. It is a block diagram of the audio | voice adjustment apparatus of Example 3. It is a flowchart which shows the adjustment step of the microphone sensitivity of Example 3, and a speaker volume.

FIG. 1 shows a basic configuration of the sound adjustment device 20 of the present invention. The voice adjustment device 20 of the present invention includes a microphone signal input unit 31 that converts an electric signal input from the microphone 10 into a voice signal, and a voice signal strength measurement that measures the strength of the voice signal and compares it with a preset voice strength threshold. The unit 32 is provided.

In addition, the sound adjusting device 20 includes a silent time measuring unit 33 that measures a silent time during which a silent state in which the voice signal is smaller than the voice strength threshold continues, and a voiced state in which the voice signal is greater than the voice strength threshold. A voiced time measuring unit 34 for measuring the sound time and a voice adjusting unit 40 for adjusting the voice signal by comparing the silent time or the voiced time with a preset time threshold value are provided.

Normally, the time that humans can speak continuously by conversation is considered to be about 5 seconds at most because of the restriction of breathing, and it is almost impossible to continue speaking without breathing for about 15 seconds. Therefore, in the audio adjusting device 20 of the present invention, when the audio signal continues in a state (sounding state) larger than the predetermined audio intensity threshold for an arbitrary fixed time (5 to 10 seconds), the audio The signal is regarded not as human speech but ambient noise, and the speech signal is adjusted so as to reduce the influence of such noise on speech recognition.

Further, in the audio adjusting device 20 of the present invention, when the audio signal continues for a certain time (20 to 30 seconds) in a state (silence state) smaller than a predetermined audio intensity threshold, the audio signal Does not include ambient noise, and adjusts the audio signal so that the speech recognition rate is improved. The voice adjustment device 20 of the present invention can improve the voice recognition rate of a device having a voice recognition function, for example, by adjusting the voice signal.

(Example 1)
FIG. 2 shows a configuration of the sound adjustment device 21 according to the first embodiment. The voice adjustment device 21 according to the first embodiment is a voice adjustment device mounted on a robot or an information terminal having a voice output function such as a voice recognition function or voice synthesis, and a microphone according to the magnitude of ambient noise in the usage environment. The voice recognition function is improved by adjusting the sensitivity (input gain).

As shown in FIG. 2, the sound adjustment device 21 receives a microphone signal input unit 31 that inputs an electrical signal from the microphone 10 and converts it into a sound signal, and measures the strength of the sound signal and compares it with a preset sound intensity threshold. And an audio signal intensity measuring unit 32.

In addition, the sound adjustment device 21 includes a silence time measuring unit 33 that measures a silence period in which a silence state in which the sound signal is smaller than the sound intensity threshold continues, and a sound condition in which the sound signal is greater than the sound intensity threshold continues. A sound duration measuring unit 34 for measuring the sound duration is provided.

Furthermore, the voice adjustment device 21 compares the silent time or the voiced time with a preset time threshold as the voice adjustment unit 40 and adjusts the sensitivity (input gain) of the microphone signal input unit 31 based on the comparison result. An adjustment unit 41 is provided.

A microphone 10 that converts external sound into an electrical signal is connected to the sound adjustment device 21, and a sound signal that has been digitized via the microphone signal input unit 31 is passed to the sound signal intensity measurement unit 32. .

The audio signal intensity measuring unit 32 has, for example, a preset audio intensity threshold value, and compares the audio intensity threshold value with the current audio signal value to determine the presence or absence of audio.

Note that the voice signal strength measurement unit 32 may be difficult to set the voice strength threshold by being incorporated in a voice recognition processing unit such as a commercial product. In this case, the processing status of the voice recognition processing unit is acquired. The presence or absence of an audio signal may be determined.

FIG. 3 is an explanatory diagram for adjusting the microphone sensitivity (input gain) when the audio signal intensity measurement unit 32 determines that the audio signal is silent. When the audio signal intensity measuring unit 32 determines that the current audio signal can be regarded as silent, the silent time measuring unit 33 measures the silent time (elapsed time in the silent state).

When the silent time continues beyond a preset silent time threshold, the microphone sensitivity of the microphone 10 is increased by the microphone sensitivity adjustment unit 41. The silent time threshold is specifically about 10 to 60 seconds, and the microphone sensitivity increase rate is preferably about 1 to 5%. The microphone sensitivity adjustment unit 41 adjusts the microphone sensitivity gradually in a quiet environment by setting the silent time threshold and the microphone sensitivity increase rate as described above.

Specifically, as shown in FIG. 3, when the silent time lasts for an arbitrary silent time threshold (T11), the microphone sensitivity is increased by several percent. If the silent time further continues from that time by the silent time threshold (T12), the microphone sensitivity is increased by several percent again. This is repeated throughout the silence period.

Note that the silence time threshold (T11) and the silence time threshold (T12) are not necessarily equal. Further, it is not necessary to make the rate of increase of each microphone sensitivity equal.

As the microphone sensitivity rises, the silence time will not continue for a long time, and the microphone sensitivity will settle down to a certain level suitable for the usage environment.

FIG. 4 is an explanatory diagram for adjusting the microphone sensitivity (input gain) when the audio signal intensity measurement unit 32 determines that the audio signal is sound. When the sound signal intensity measuring unit 32 determines that the current sound signal can be regarded as sound, the sound time measuring unit 34 measures the sound time (elapsed time of the sound state).

When the sound duration continues beyond an arbitrary sound duration threshold set in advance, the microphone sensitivity of the microphone 10 is lowered by the microphone sensitivity adjustment unit 41. The sound duration threshold is specifically about 5 to 20 seconds, and the rate of decrease in sensitivity is preferably about 10 to 50%, and is usually set based on the time during which a person can speak with one breath. The microphone sensitivity adjustment unit 41 adjusts the microphone sensitivity of the microphone to be quickly lowered under a noisy environment by setting the sound duration threshold and the microphone sensitivity increase rate as described above.

Specifically, if the sound duration lasts for an arbitrary sound duration threshold (T21), the microphone sensitivity is reduced by several percent to several tens of percent. Further, if the sound duration continues from that time for the sound duration threshold (T22), the microphone sensitivity is lowered again by several percent to several tens percent. This is repeated throughout the duration of the sound.

Note that the sound time threshold (T21) and the sound time threshold (T22) are not necessarily equal. Further, it is not necessary to make the rate of decrease of the input gain equal in each.

As the microphone sensitivity decreases, the sound duration will not continue as long as before, and the microphone sensitivity will settle to a certain level suitable for the usage environment.

FIG. 5 shows a processing flow for adjusting the microphone sensitivity (input gain) in the sound adjustment device 21 of the first embodiment. The processing flow will be described with reference to FIG.
S11. Start the audio adjustment process.
S12. In the audio signal intensity measurement unit 32, the intensity of the audio signal is compared with an arbitrary audio intensity threshold (or based on whether the audio is being recognized), and if the intensity of the audio signal is greater than the audio intensity threshold, If the intensity of the audio signal is smaller than the audio intensity threshold, it is determined that there is no sound.
S13. When the sound signal intensity measuring unit 32 determines that there is no sound (no sound signal), the soundless time measuring unit 33 measures the soundless time (the duration of the section without the sound signal).
S14. If the silent time is shorter than an arbitrary silent time threshold, the process returns to the start of the sound adjustment processing (S11) without changing the microphone sensitivity. If the silent time is longer than the silent time threshold, the process proceeds to the next process (S15).
S15. When the silent time is longer than the silent time threshold, it is determined that there is a margin in microphone sensitivity in a quiet environment with no noise around the microphone, and the microphone sensitivity of the microphone signal input unit 31 is increased by an arbitrary ratio. The process returns to the start of the sound adjustment process (S11).
S16. When the voice signal strength measuring unit 32 determines that there is sound (there is a voice signal), the voiced time measuring unit 34 measures the voiced time (the duration of a section where the voice signal is present).
S17. If the sound duration is shorter than an arbitrary sound duration threshold, it is determined that the conversation is a normal conversation, and the process returns to the start of the sound adjustment processing (S11) without changing the microphone sensitivity. If the sound time is longer than the sound time threshold, the process proceeds to the next process (S18).
S18. If the sound time is longer than the sound time threshold, it is determined that there is constant noise around the sound, and the microphone sensitivity of the microphone 10 is lowered by an arbitrary ratio by the microphone signal input unit 31, and then the sound adjustment processing Return to the start (S11).

As in the above processing flow, the present invention is characterized in that the length of the silence state and the sound state is measured from the voice signal and adjusted to a voice signal suitable for voice recognition. That is, according to the sound adjustment device 21 of the first embodiment, the ambient noise and the human conversation are distinguished by the length of time during which the silent time or the sound time continues, and the optimum microphone sensitivity is obtained according to the use environment. Can be set.

(Example 2)
FIG. 6 shows the configuration of the audio adjustment device 22 according to the second embodiment. The audio adjusting device 22 according to the second embodiment includes a speaker volume adjusting unit 42 and a speaker signal output unit 43 as the audio adjusting unit 40, and the volume of the speaker 50 (depending on the usage environment such as a quiet environment or a noisy environment with a lot of noise). This is an audio adjustment device that adjusts the output gain easily.

In the audio adjustment device 22, an audio signal is input through the microphone 10 and the microphone signal input unit 31, the intensity of the audio signal is measured by the audio signal intensity measurement unit 32, and the silent time measurement unit 33 and the sound time measurement unit 34 are measured. The process until the silent time and the voiced time are measured is the same as that of the sound adjustment device 21 of the first embodiment.

After that, in the sound adjustment device 22 of the second embodiment, when the silence time measurement unit 33 determines that the silence time is longer than an arbitrary silence time threshold set in advance, the speaker volume adjustment unit 42 controls the volume of the speaker 50. Reduce (output gain). Specifically, the silent time threshold is preferably about 10 to 60 seconds, and the volume decrease rate is preferably about 1 to 5%. The speaker volume adjustment unit 42 adjusts the volume of the speaker 50 to gradually decrease in a quiet environment by setting the silent time threshold and the volume decrease rate.

Further, the speaker volume adjustment unit 42 increases the volume (output gain) of the speaker 50 by the speaker volume adjustment unit 42 when the sound duration is longer than an arbitrary sound duration threshold set in advance. The sound duration threshold is specifically about 5 to 20 seconds, and the rate of increase in volume is preferably about 10 to 50%, and is usually set based on the time during which a person can speak with one breath. The speaker volume adjustment unit 42 adjusts the volume of the speaker 50 to be quickly increased in a noisy environment by setting the silent time threshold and the volume decrease rate.

FIG. 7 shows a processing flow for adjusting the speaker volume (output gain) in the sound adjustment device 22 of the second embodiment. The processing flow will be described with reference to FIG.
S11. Start the audio adjustment process.
S12. In the audio signal intensity measurement unit 32, the intensity of the audio signal is compared with an arbitrary audio intensity threshold (or based on whether the audio is being recognized), and if the intensity of the audio signal is greater than the audio intensity threshold, If the intensity of the audio signal is smaller than the audio intensity threshold, it is determined that there is no sound.
S13. When the sound signal intensity measuring unit 32 determines that there is no sound (no sound signal), the soundless time measuring unit 33 measures the soundless time (the duration of the section without the sound signal).
S14. If the silent time is shorter than an arbitrary silent time threshold, the process returns to the start of the sound adjustment processing (S11) without changing the volume of the speaker. If the silent time is longer than the silent time threshold, the process proceeds to the next process (S21).
S21. If the silent time is longer than the silent time threshold, it is determined that the environment is quiet and there is no noise, the volume (output gain) of the speaker 50 is lowered at an arbitrary rate, and then the voice adjustment process is started. Return to (S11).
S16. When the voice signal strength measuring unit 32 determines that there is sound (there is a voice signal), the voiced time measuring unit 34 measures the voiced time (the duration of a section in which the voice signal is present).
S17. If the sound duration is shorter than the arbitrary sound duration threshold, it is determined that the conversation is normal, and the flow returns to the start of the sound adjustment processing (S11) without changing the output gain. If the sound time is longer than the sound time threshold, the process proceeds to the next process (S22).
S22. If the sound time is longer than the sound time threshold, it is determined that the sound is ambient noise, the sound volume (output gain) of the speaker 50 is increased by an arbitrary ratio, and then the sound adjustment process is started (S11). Return to.

According to the voice adjustment device 22 of the second embodiment, since the volume (output gain) of the speaker 50 is adjusted according to the intensity of ambient noise, the volume of human conversation is easily adjusted according to the usage environment. be able to.

(Example 3)
FIG. 8 shows the configuration of the sound adjustment device 23 according to the third embodiment. The sound adjustment device 23 according to the third embodiment is a combination of the sound adjustment device 21 according to the first embodiment and the sound adjustment device 22 according to the second embodiment. The voice adjustment device 23 according to the third embodiment adjusts the sensitivity (input gain) of the microphone so that the voice recognition function is in a good state in a quiet environment or a noisy environment where there is a lot of noise. This is an audio adjustment device that adjusts the volume (output gain) of a speaker so that it can be easily heard.

As shown in FIG. 8, the voice adjustment device 23 of the third embodiment has the configuration of the voice adjustment device 21 of the first embodiment and the configuration of the voice adjustment device 22 of the second embodiment. The details of each component of the sound adjustment device 23 are the same as those in the first and second embodiments, and thus the description thereof is omitted.

FIG. 9 shows a processing flow of sound adjustment of the microphone and the speaker by the sound adjustment device 23 of the third embodiment. The processing flow will be described with reference to FIG.
S11. Start the audio adjustment process.
S12. In the audio signal intensity measurement unit 32, the intensity of the audio signal is compared with an arbitrary audio intensity threshold (or based on whether the audio is being recognized), and if the intensity of the audio signal is greater than the audio intensity threshold, If the intensity of the audio signal is smaller than the audio intensity threshold, it is determined that there is no sound.
S13. When the sound signal intensity measuring unit 32 determines that there is no sound (no sound signal), the soundless time measuring unit 33 measures the soundless time (the duration of the section without the sound signal).
S14. If the silent time is shorter than an arbitrary silent time threshold, the process returns to the start of the sound adjustment processing (S11) without changing the microphone sensitivity and the speaker volume. If the silent time is longer than the silent time threshold, the process proceeds to the next process (S15).
S15. If the silent time is longer than the silent time threshold, it is determined that the microphone sensitivity is sufficient in a quiet environment with no surrounding noise, and the microphone sensitivity (input sensitivity) is increased by an arbitrary ratio. After lowering the speaker volume (output gain) at an arbitrary rate, the process returns to the start of the audio adjustment process (S11).
S16. When the voice signal strength measuring unit 32 determines that there is sound (there is a voice signal), the voiced time measuring unit 34 measures the voiced time (the duration of a section in which the voice signal is present).
S17. If the sound duration is shorter than an arbitrary sound duration threshold, it is determined that the conversation is normal, and the process returns to the start of the sound adjustment processing (S11) without changing the microphone sensitivity and the speaker volume. If the sound time is longer than the sound time threshold, the process proceeds to the next process (S18).
S18. If the sound time is longer than the sound time threshold, it is determined that there is constant noise around the sound, and the microphone sensitivity (input sensitivity) is reduced by an arbitrary ratio. ) At an arbitrary rate, and then the process returns to the start of the audio adjustment process (S11).

According to the voice adjustment device 23 of the third embodiment, the ambient noise and the human conversation are distinguished based on the length of time during which the silent time or the voiced time lasts. Since the speaker volume is set, it is possible to easily adjust the volume of human conversation according to the usage environment. In addition, by providing the voice adjustment device 23 of the present invention, it is possible to improve the voice recognition rate in a device having a voice recognition function.

According to the present invention, it can be used in any device having a voice recognition function and a voice response function.

DESCRIPTION OF SYMBOLS 10

Microphone

20, 21, 22, 23 Audio | voice adjustment apparatus 31 Microphone signal input part 32 Audio | voice signal strength measurement part 33 Silent time measurement part 34 Sound time measurement part 40 Audio | voice adjustment part 41 Microphone sensitivity adjustment part 42 Speaker volume adjustment part 43 Speaker Signal output unit 50 Speaker

Claims

A microphone signal input unit that converts an electrical signal input from the microphone into an audio signal;
A voice signal strength measurement unit that measures the strength of the voice signal and compares it with a preset voice strength threshold;
A silent time measuring unit for measuring a silent time during which a silent state in which the voice signal is smaller than the voice intensity threshold is continued;
A sound duration measuring unit that measures a sound duration in which a sound state in which the voice signal is greater than the voice intensity threshold continues;
An audio adjusting device comprising: an audio adjusting unit that adjusts the audio signal by comparing the silent time or the sounded time with a preset time threshold value.
The voice adjustment unit includes a microphone sensitivity adjustment unit for adjusting an input gain of the microphone signal input unit,
The input gain is increased when the silent time is longer than a preset silent time threshold, and the input gain is decreased when the voiced time is longer than a preset silent time threshold. The sound adjustment device described in 1.
The audio adjustment unit includes a speaker signal output unit that outputs an audio signal to a speaker, and a speaker volume adjustment unit for adjusting an output gain of the speaker signal output unit,
2. The output gain is lowered when the silent time is longer than a preset silent time threshold, and the output gain is increased when the voiced time is longer than a preset silent time threshold. The sound adjustment device described in 1.
The audio adjustment unit adjusts an output gain of the microphone signal input unit, a microphone sensitivity adjustment unit for adjusting an input gain of the microphone signal input unit, a speaker signal output unit that outputs an audio signal to a speaker, and the speaker signal output unit. Speaker volume adjustment unit,
Increasing the input gain and decreasing the output gain when the silence time is longer than a preset silence time threshold,
The sound adjustment device according to claim 1, wherein when the sound duration is longer than a preset sound duration threshold, the input gain is decreased and the output gain is increased.