JPH10224897A - Processing circuit for audio signal - Google Patents

Abstract

PROBLEM TO BE SOLVED: To facilitate listening of even continuous sounds in speech while reducing the effect of masking at the time of relay. SOLUTION: A variable gain amplifier 12 is provided on the signal line of audio signal S11. A detection circuit 23 for detecting the end point of audio signal S11 and a control circuit 24 for controlling the gain of variable gain amplifier 12 according to the detection output of this detection circuit 23 are provided. When the end point of audio signal S11 is detected by the detection circuit 23, at the variable gain amplifier 12, the amplitude of audio signal S11 is enlarged for a prescribed period by the control output of control circuit 24.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

The present invention relates to a hearing aid, a telephone,
The present invention relates to an audio signal processing circuit used in fields such as a loudspeaker and an audio communication.

[0002]

2. Description of the Related Art When transmitting or reproducing sound, if there is much reverberation or echo in the transmission system or reproduction system, the clarity of the resulting sound is reduced. Therefore, in such a case, processing such as slowing down the utterance speed, finely decomposing a continuously uttered speech sound, and reproducing it at a later time is performed.

When a high frequency such as a consonant is difficult to hear, the high frequency may be emphasized by a frequency equalizer process. Furthermore, so-called successive masking (when energetic vowels and consonants continue,
Attempts have also been made to apply a weighting function taking into account the phenomenon of consonants being masked by the vowels).

[0004] Further, the above processing may be performed for a hearing-impaired person or an elderly person.

[0005]

However, as described above, if the utterance speed is reduced or a speech sound that is continuously uttered is decomposed, the following problems occur.

[0006] 1. A time lag occurs between the original sound and the original sound. Therefore, it cannot be used for conversations. In addition, even if you listen to a broadcast, the time it takes to finish listening will be long. 2. Since the rate of change of the voice component is also an important clue in the perception judgment of the speech sound, if the speech speed is reduced, the clue may change and be perceived as another speech sound. 3. If a speech is decomposed and played slowly, information as a unit of speech or information of a transitional change may be lost, resulting in poor clarity. 4. A sound with a constantly amplified high frequency may be uncomfortable due to a loss of timbre, or may be difficult to hear. 5. In the process of applying the weighting function in consideration of the successive masking, at least the time length of the weighting function is delayed, and the immediacy is lost. As a result, a time lag may occur between the movement of the mouth and the processed sound, which may adversely affect the clarity. Also, when there is acoustic feedback from the earphone to the microphone, a time delay causes a phenomenon such as reverberation.

[0007] The present invention aims to eliminate the above problems.

[0008]

According to the present invention, there is provided a variable gain amplifier to which an audio signal is supplied, a detection circuit for detecting an end point of the audio signal, and a detection circuit for detecting an end point of the audio signal. A control circuit for controlling the gain of the variable gain amplifier, wherein when the detection circuit detects the end point of the audio signal, the variable gain amplifier controls the amplitude of the audio signal by the control output of the control circuit. Is increased over a predetermined period of time. Therefore, when a consonant follows a continuous voice, its amplitude is enlarged.

[0009]

DESCRIPTION OF THE PREFERRED EMBODIMENTS The voice of a normal conversation is
It is said to be pronounced with a certain degree of cohesion, and it is said that the speech perception of speech is also performed from both the perception of each sound and the perception of the characteristics of a set of speech sounds.

[0010] In addition, in the case of a normal hearing person, in a place where there is little noise and the listening environment is good, the voice can be sufficiently heard without performing any special sound processing, but in a place where there is a lot of noise, words are not well heard. Or become. There are several possible causes, but the main one is the effect of successive masking. In other words, the previous vowel may mask the consonant part of the next sound, and as a result, the auditory sensitivity of the consonant becomes poor and the consonant becomes difficult to hear.

Therefore, according to the present invention, the end point of a group of voices is detected, and the amplitude of the voice signal (particularly its high-frequency component) is amplified for a period of several tens to several tens of milliseconds from this end point. Thus, the auditory sensitivity to consonants is relatively increased. Also, the end point of the united voice is detected by analyzing the level change of the pitch component and the formant component of the voice.

FIG. 1 shows an embodiment of the present invention.
The original unprocessed audio signal S11 is supplied to the variable gain amplifier 12 through the input terminal 11, and the output signal S12 of the amplifier 12 is extracted to the output terminal 13.

Further, the signal S11 at the terminal 11 is converted to a band-pass filter 21 and a level detection circuit 2 for preprocessing.
2 in turn. In this case, the bandpass filter 2
1 makes it easier to detect the end point of the continuous voice, and
The pitch component and the formant component are extracted from the signal S11 so that the influence of noise is reduced. Therefore, the pass band is, for example, 150 Hz to 1000 Hz.

The level detection circuit 22 detects the end point of the continuous sound using the output signal S21 of the band-pass filter 21. For this reason, the level detection circuit 22 forms a signal S22 indicating the level of the signal S11 by, for example, performing double-wave rectification on the signal S21 and extracting a low-frequency component (for example, a component of 60 Hz or less).

The detection signal S22 of the level detection circuit 22 is supplied to an end point detection circuit 23, where the end point of the continuous sound is detected, and the detection signal S23 of the end point is supplied to a gain control circuit 24. A control signal S24 is formed, and this signal S24 is supplied to the variable gain amplifier 12 as a gain control signal.

In this case, the end point of the continuous voice is detected when the voice level, that is, the level of the signal S22 once exceeds the first threshold value (voice presence determination threshold value), and then the second threshold value. It is performed when it becomes smaller than a threshold value (speech end determination threshold value). Further, the gain is controlled, for example, from several tens of msec to several tens m
The gain is increased over a period of seconds. However, the first threshold is equal to or greater than the second threshold.

According to such a configuration, while the audio signal S11 input to the terminal 11 is continuous, the control signal S11
By 23, the gain of the variable gain amplifier 12 is fixed to the reference value. Therefore, the audio signal S11 is directly taken out to the terminal 13 as the output signal S12.

However, when the continuation of the audio signal S11 is completed, the gain of the variable gain amplifier 12 is made larger than the reference value by the control signal S24 from the time of the end point. If there is the audio signal S11, the amplitude of the audio signal S12 becomes larger than the original amplitude. Therefore, even if the perceptual sensitivity to the consonant immediately after the continuous voice (voice signal S11) is reduced by successive masking, the amplitude of the consonant is larger than the original size, and the sensitivity is reduced. Are canceled out, and the clarity of the voice including the consonant is improved.

2 to 4 show an embodiment of a method in which the detection circuit 23 and the control circuit 24 form the control signal S24 from the detection signal S22. That is, in this case, all of the circuits shown in FIG. 1 are digitized and configured by, for example, a DSP. Then, the audio signal S11 is a digital audio signal obtained by A / D converting the original analog audio signal before processing.

The detection circuit 23 and the control circuit 24
2, the processing routine 100 of FIG. 2 is executed for each sample of the digital audio signal S11. In this routine 100, when the gain of the amplifier 12 is changed, the level of the audio signal S12 is changed gradually as shown in FIG. 5, for example, in order to prevent discontinuity.

In the routine 100 and the following description, the meaning of each variable is as follows.

E (i): the level of the i-th sample of the audio signal S11. peak: peak value of the audio signal S11. peakmax: Value for limiting the upper limit of the value peak. rate: The rate of what percentage of the signal S11 becomes smaller than the value of peak to determine that the sound is gone. voicemin: First threshold (voice presence determination threshold). When the signal S11 exceeds this value, the presence of voice is recognized. threshold: The second threshold value (voice end determination threshold value). When the signal S11 becomes smaller than this value, it is determined that the sound is the end point. voicemin ≧ threshold, th
reshold = peak × rate. slope1: period T during which the gain of the amplifier 12 is increased
Length of 1. slope2: The length of the period T2 in which the gain of the amplifier 12 is restored. enable: Flag indicating whether or not voice exists in the (i-1) th sample. When enable = "0", no voice exists. When enable = "1", voice is present. jj: A variable for counting time from the end point of a continuous voice.

In the routine 100, first, at step 101, it is determined whether or not e (i) <threshold holds, thereby determining whether or not there is no voice in the i-th sample. If it does not exist (e (i) <threshold), the process proceeds to step 102. Then, in step 102, by checking whether enable = "1", it is determined whether or not the voice exists in the (i-1) -th sample which is the immediately preceding sample. If a voice exists in the previous sample (when enable = "1"), the process proceeds to step 103.

Therefore, the process proceeds to step 103 when the sound is present in the preceding sample and the sound is not present in the succeeding sample of the two consecutive samples. This is when the sound that has been played ends. That is, when the end of the continuous voice is detected.

Therefore, in step 103, the following
In preparation for the (i + 1) -th sample, enable = "0" is set and jj = 0, and the processing for the i-th sample is completed.

In step 102, when the signal S11 does not exist in the immediately preceding sample (enable =
If “0”), the process proceeds to step 111, where jj
It is determined whether or not the sample time is included in the period T1 by checking whether or not <slope1. If the sample time is included (when jj <slope1), the process proceeds to step 112. That is, the process proceeds to step 112 during the period T1.

Then, in step 112, the gain of the amplifier 12 is increased by one step. Also, the variable jj is incremented by “1”. Then, the processing for the i-th sample is completed.

Further, in step 111, when the sampling time is not included in the period T1 (when jj ≧ slope1), the processing proceeds to step 121. That is, when the period T1 has passed, the process proceeds to step 121.

Then, in step 121, jj
By checking whether or not ≧ slope1 and jj <(slope1 + slope2), it is determined whether or not the sample time is included in the period T2.
If e1 and jj <(slope1 + slope2), the process proceeds to step 122. That is, the process proceeds to step 122 during the period T2.

In this step 122, the gain of the amplifier 12 is reduced by one step. Also, the variable jj is incremented by “1”. Then, the processing for the i-th sample is completed.

Further, in step 121, when the sampling time has passed the period T2 (jj ≧ (slope1 + sl
In the case of ope2), the processing for the i-th sample is ended without changing the gain of the amplifier 12.

Thus, according to the above-described processing, when the continuous voice ends, this is detected, and the gain of the amplifier 12 is controlled as shown in FIG.

In the routine 100, an upper limit is set on the peak value of the audio level to prevent the peak value of the audio level from becoming extremely large when the gain of the amplifier 12 is increased. In the routine 100, the second threshold value is set for each continuous voice, and a peak of the voice level is found after the voice level exceeds the first threshold value, and several% of the peak value is found. It is set to a value of ~ several tens%.

That is, in step 101, when a voice is present (when e (i) ≧ threshold), the process proceeds to step 131. And this step 131
In step (1), it is determined whether or not enable = "0" to determine whether or not voice exists in the (i-1) th sample which is the immediately preceding sample.
When there is no sound in the previous sample (enable =
If “0”, the process proceeds to step 141.

In this case, since the process proceeds to step 141 through steps 101 and 131, this means that the sound does not exist in the preceding sample and the succeeding sample in the two consecutive samples does not exist. This is the case when audio is present, that is, when audio is started.

Therefore, in step 141, e (i)>
A check is made to see if the voice has been started by checking if it is voicemin, and if it has been started (e (i)> voicemin), the process proceeds to step 1
Proceeding to 42, it is determined whether the level e (i) exceeds the upper limit value peakmax. And when it does not exceed (e
(i) ≦ peakmax), the process proceeds to step 143, where the peak value peak is set to the level e (i), and thereafter,
The process proceeds to step 145.

In step 142, the level e
When (i) exceeds the upper limit value peakmax (e (i)> peakmax
), The process proceeds to step 144, where the peak value peak is set to the upper limit value peakmax, and then the process proceeds to step 145.

Then, in step 145, the second threshold value threshold is set to the threshold value th according to the peak value peak set in step 143 or 144.
reshold = set to the value indicated by peak x rate and enable = for the next (i + 1) th sample
It is set to "1", and thereafter, the process proceeds to step 161.

Further, in step 141, when voice has not been started (when e (i) ≦ voicemin),
The process proceeds directly from step 141 to step 161 and the peak value peak and the flag enable are not set.

Thus, when the sound is started, the second threshold value threshold is set corresponding to the level at the start.
Is set.

On the other hand, if it is determined in step 131 that a sound exists in the immediately preceding sample (when enable = "1"), the process proceeds to step 151. in this case,
Since the process proceeds to step 151 through steps 101 and 131, this is when there is speech in both two consecutive samples,
This is when the sound is continuous. In other words, this is a period during which the sound is continuous.

Then, in step 151, the level e (i) of the i-th sample is compared with the previous peak value peak, and when the level e (i) is larger (e (i)> peak)
), The processing proceeds from step 151 to step 15
Proceeding to 2, it is determined whether or not the level e (i) exceeds the upper limit value peakmax. And when it does not exceed (e (i)
(<Peakmax), the process proceeds to step 153, where the peak value peak is updated to the level e (i), and then the process proceeds to step 155.

In step 152, level e
When (i) exceeds the upper limit peakmax (e (i) <peakmax
), The process proceeds to step 154, where the peak value peak is updated to the upper limit value peakmax, and then the process proceeds to step 155.

Then, in step 155, the second threshold value threshold is set to th according to the peak value peak updated in step 153 or step 154.
reshold = peak × rate is updated to the value indicated, and then the process proceeds to step 161.

Further, in step 151, when the peak value peak so far is larger (when e (i) ≦ peak), the process proceeds from step 151 to step 161 as it is, and the peak value peak is not updated.

Thus, when the sound is continuous,
A peak value peak in the continuous period is found, and a second threshold value th according to the peak value peak is obtained.
reshold is updated.

Then, when the processing proceeds to step 161,
By checking whether jj> 0 and jj <slope1, it is determined whether or not the sample time is included in the period T1, and when it is included (jj> 0 and jj <slop1).
In the case of e1, the process proceeds to step 162. That is, the process proceeds to step 162 during the period T1.

Then, in this step 162, the gain of the amplifier 12 is increased by one step. Also, the variable jj is incremented by “1”. Then, the processing for the i-th sample is completed.

Further, in step 161, when the sampling time is not included in the period T 1 (when jj ≧ slope 1), the processing proceeds to step 171. That is, when the period T1 has passed, the process proceeds to step 171.

Then, in this step 171, jj
By checking whether or not ≧ slope1 and jj <(slope1 + slope2), it is determined whether or not the sample time is included in the period T2.
If e1 and jj <(slope1 + slope2), the process proceeds to step 172. That is, the process proceeds to step 172 during the period T2.

Then, in this step 172, the gain of the amplifier 12 is reduced by one step. Also, the variable jj is incremented by “1”. Then, the processing for the i-th sample is completed.

Further, in step 171, when the sampling time has passed the period T2 (jj ≧ (slope1 + sl
In the case of (ope2)), the process proceeds to step 181, where the variable jj is initialized to "0", and the process for the i-th sample is completed.

By the above-described processing, the audio signal S11 is output for a period of about several tens of milliseconds to several tens of milliseconds starting from the end point of the continuous audio.
Therefore, even if the auditory sensitivity to the consonant immediately after the continuous sound (sound signal S11) is reduced by successive masking, the decrease in the sensitivity is caused by the increase in the amplitude of the signal S11. As a result, the intelligibility of the voice including the consonant is improved.

FIG. 4 shows the result of the observation of the sound waveform. FIG. 4A shows the waveform of the sound signal S11 which has not been processed by the routine 100, and FIG. 4B shows the waveform of the sound signal S12 which has been processed by the routine 100. is there. The utterance content at this time is "Please write on the first line".

As shown by arrows A, E and F, when the time interval from the end point of a continuous voice to the next consonant is short, successive masking for the consonant is large, but the amplitude of the consonant is enlarged. Being emphasized.
As shown by arrows B, C, and D, when the time interval to the next consonant is long, successive masking is small, but the amplitude of the consonant is not emphasized correspondingly.

Therefore, according to the above-described processing circuit, the following effects can be obtained when the sound is transmitted or reproduced in a certain system such as reverberation or echo, or when a hearing-impaired person or an elderly person hears the sound. . 1. Since only the consonant is emphasized so that successive masking to the next consonant is reduced, the sound is clear and the clarity can be improved. 2. When emphasizing the high frequency of the voice is always accompanied by an unpleasant sensation such as a loss of timbre balance, the consonant is emphasized only when successive masking is occurring, so there is no such unpleasant sensation. 3. Since the processing can be performed immediately in principle, there is no time difference between the movement of the mouth of the speaker and the processed sound. Also, even if there is acoustic feedback from the earphone to the microphone, it does not sound like reverberation, so it is easy to hear. 4. The rate of change of the speech component, which is important for the perceptual judgment of the speech sound, the information as a unit of the speech sound, and the information of the transient change portion are not lost.

In the above description, the amount of masking by successive masking varies depending on the level and duration of a sound portion (continuous sound section) acting as a disturbing sound, and the time from the end point of the continuous sound portion. The period during which the amplitude of the original audio signal S11 is controlled and its magnitude can be adjusted according to the masking amount of the successive masking. When the masking amount is large, the amplitude and the control period can be increased. Good.

In the above description, the variable gain amplifier 1
In 2, the amplitude is increased in the entire band of the audio signal S11, but the amplitude may be increased only in the high band corresponding to the consonant.

[0059]

According to the present invention, the sound is clear,
Clarity can be improved. Also, there is no unpleasant sensation as when always emphasizing the high frequency range of the sound. Furthermore, there is no time difference between the movement of the mouth of the speaker and the processed sound.

Also, even if there is acoustic feedback from the earphone to the microphone, it does not look like a reverberant sound and is easy to hear. Further, effective information for perceptual judgment of speech sounds is not impaired.

[Brief description of the drawings]

FIG. 1 is a system diagram illustrating one embodiment of the present invention.

FIG. 2 is a flowchart showing a part of one embodiment of the present invention.

FIG. 3 is a flowchart showing a continuation of FIG. 2;

FIG. 4 is a flowchart showing a continuation of FIG. 3;

FIG. 5 is a diagram for explaining the present invention.

FIG. 6 is a diagram for explaining the present invention.

[Explanation of symbols]

12: variable gain amplifier, 21: band-pass filter, 2
2: Level detection circuit, 23: End point detection circuit, 24: Gain control circuit, 100: Processing routine

──────────────────────────────────────────────────続 き Continued on front page (51) Int.Cl. ⁶ Identification code FI H03G 3/00 H03G 3/00 A

Claims (7)

[Claims] A variable gain amplifier to which an audio signal is supplied; a detection circuit for detecting an end point of the audio signal; and a control circuit for controlling a gain of the variable gain amplifier according to a detection output of the detection circuit. A variable gain amplifier configured to increase the amplitude of the audio signal over a predetermined period by a control output of the control circuit when the end point of the audio signal is detected by the detection circuit. Processing circuit. 2. The audio signal processing circuit according to claim 1, wherein the detection circuit terminates the audio after the level of the audio signal once exceeds a first threshold for recognizing the presence of audio. The point in time when the value becomes smaller than the second threshold value determined as
An audio signal processing circuit configured to generate the detection output. 3. The audio signal processing circuit according to claim 1, wherein the audio signal is supplied to a band-pass filter to extract a pitch component and a formant component of the audio as an output signal. 3. The audio signal processing circuit according to claim 1, wherein an end point of said audio signal is detected from an output signal of said band pass filter. 4. The audio signal processing circuit according to claim 1, wherein the first threshold value is equal to or larger than the second threshold value. 5. The audio signal processing circuit according to claim 1, wherein the second threshold value is set separately from a maximum value of the level of the audio signal during a period in which the audio exists. An audio signal processing circuit that is set to a few percent to a few tens percent of the smaller of the upper limit and the upper limit. 6. The audio signal processing circuit according to claim 1, wherein a gain of said variable gain amplifier is controlled by a level of said continuous audio portion. 7. The audio signal processing circuit according to claim 1, wherein a gain of said variable gain amplifier is controlled by a time from an end point of said audio signal.