JPH1078798A - Voice signal processor - Google Patents

Abstract

PROBLEM TO BE SOLVED: To clearly catch a conversational voice by calculating an absolute value of a difference of a sample value and deciding a coefficient according to the absolute value of the calculated difference. SOLUTION: A signal processing means 3 is constituted of a difference extraction means 5 calculating the difference of the sample values adjacent to each other of output signals of a time frame division means 2, that is, continuous sample values and a signal expansion/reduction means 6 expanding or reducing the output signal of an outside sound sampling means 1 based on the absolute value of the difference of the difference extraction means 5. Then, when the mean value of the absolute value of the difference exceeds a prescribed threshold value, all sample values in a time frame are multiplied by the same coefficient so that the maximum value among the sample values incorporated in the time frame becomes the prescribed value. Thus, a consonant is separated from an environmental noise, and only the level of the consonant is amplified. That is, since the conversational voice is separated from the stationary environmental noise by simple signal processing to be amplified selectively, the conversational voice is made be easy to catch.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an audio signal processing apparatus applied to a hearing aid or the like for listening to conversational sounds with a high listening clarity.

[0002]

2. Description of the Related Art Conventionally, a hearing impaired person who uses a hearing aid such as a hearing aid often hears a high-level sound as well as a normal person, although a low-level sound cannot be heard. Conversational sounds include vowels and consonants, and consonants (especially unvoiced consonants such as / s / and / p /) are low in level and are difficult for hearing impaired people to hear.

[0003] In addition, hearing has a phenomenon called masking, and a weak consonant tends to be harder to hear because it is hidden by a strong vowel. A person with normal hearing can hear consonants even with masking because they can hear low-level sounds well,
A hearing impaired person has difficulty hearing consonants due to masking.

If consonants cannot be heard, it is difficult to hear conversation sounds clearly. In order to solve such inconveniences, for example, a hearing aid device as disclosed in Japanese Patent Publication No. 8-34652 has been proposed. In this hearing aid device, a technique is used in which the amplification characteristic is changed according to the spectrum of the input sound to match the hearing characteristics of a hearing-impaired person.

[0005]

However, even if such a hearing aid is used, if the amplification for a low-level sound is increased beyond a certain level, the ambient noise also increases at the same time, so that only the consonant level can be sufficiently reduced. Cannot be amplified,
The hearing aid wearer was unable to hear speech clearly.

SUMMARY OF THE INVENTION The present invention has been made in view of such problems of the prior art, and an object of the present invention is to provide an audio signal processing apparatus capable of clearly hearing conversation sounds. Is what you do.

[0007]

In order to solve the above-mentioned problems, a first aspect of the present invention is a sampling means for sampling a speech signal at regular intervals to generate a time-series sample value, A time frame dividing means for sequentially generating a time frame of the sample value by dividing the time frame into a plurality of continuous, and a signal processing means for performing signal processing on the time frame and updating the sample value included in the time frame. Output means for converting the updated sample value into an audio signal and outputting the signal, wherein the signal processing means calculates an absolute value of a difference between two successive sample values, and And a signal enlargement / reduction means for multiplying the sample value by a coefficient determined according to the absolute value of the difference.

According to a second aspect of the present invention, in the audio signal processing apparatus according to the first aspect, the time frame dividing means changes a boundary of the sample value so that its sign changes so as to correspond to a zero crossing point of the speech signal. Is separated by

According to a third aspect of the present invention, in the audio signal processing apparatus according to the first or second aspect, the signal enlarging / reducing means is provided when the absolute value of the difference between the sample values exceeds a predetermined threshold value. Determines the coefficient so that the maximum value of the absolute value of the sample value included in the time frame becomes a constant value, and when the absolute value of the difference between the sample values does not exceed the threshold value, The coefficient is determined such that the maximum of the absolute value of the sample value included in the frame is smaller than a certain value.

According to a fourth aspect of the present invention, in the audio signal processing apparatus according to the first, second or third aspect, the difference extracting means includes:
With respect to all the sample values included in the time frame, absolute values of two consecutive differences are all calculated, and further, an average value thereof is calculated.

According to a fifth aspect of the present invention, in the audio signal processing apparatus according to the first, second, third or fourth aspect, the difference extracting means includes the first sample value and / or the last sample value included in the time frame. With respect to a sample value, an absolute value of a difference between the sample value and a continuous sample value is calculated.

[0012]

Embodiments of the present invention will be described below with reference to the accompanying drawings. Here, FIG. 1 is a block diagram showing a conceptual configuration of an audio signal processing device according to the present invention, and FIG.
3 is a configuration diagram of the audio signal processing device, FIG. 3 is a flowchart showing a processing procedure of the audio signal processing device, FIG. 4 is a flowchart showing a procedure of time frame division, FIG. 5 is a flowchart showing a procedure of difference calculation, and FIG. 7 is a waveform diagram showing a signal amplified by the audio signal processing device, FIG. 8 is a diagram showing an input waveform (a) and a waveform (b) after processing, and FIG. FIG. 10 is a decision diagram showing an example of the effect of a hearing aid to which the audio signal processing device according to the present invention is applied, FIG. 10 is a flowchart showing a procedure of another embodiment of time frame division, and FIG. 11 is another embodiment of difference extraction. 6 is a flowchart showing the procedure of FIG.

[0013] The audio signal processing device, as shown in FIG.
External sound sampling means 1 for inputting and sampling (digital sampling) an external sound, time frame dividing means 2 for dividing a signal (sample value) sampled by the external sound sampling means 1 into time frames, A signal processing unit 3 for performing an arithmetic processing on an output signal of the time frame dividing unit 4 based on a predetermined program, and an output unit 4 for converting a signal processed by the signal processing unit 3 into an analog signal and then outputting the analog signal as an acoustic signal. It is configured.

The external sound sampling means 1 comprises a microphone 11 for inputting external sound and converting it into an electric signal, and an A / D converter 12 for converting an analog signal output from the microphone 11 into a digital signal.

Further, the signal processing means 3 includes an output signal of the time frame dividing means 2, that is, a difference extracting means 5 for calculating a difference between adjacent sample values of successive sample values, and an output signal of the difference extracting means 5. It comprises a signal enlarging / reducing means 6 for enlarging or reducing the output signal of the external sound sampling means 1 based on the (absolute value of the difference).

The output unit 4 converts a digital signal output from the signal processing unit 3 into an analog signal, and converts an electric signal output from the D / A converter 31 into an acoustic signal. It consists of an earphone 32.

The time frame dividing means 2 and the signal processing means 3 are composed of an arithmetic unit 7 as shown in FIG. The arithmetic unit 7 includes a CPU (arithmetic processing unit) 21 for performing arithmetic processing based on a predetermined program, and R as storage means for storing programs and constants.
An OM (Read Only Memory) 22 and a RAM (Random Access Memory) 23 which is a storage unit for storing data and variables are provided.

The processing procedure by the audio signal processing device configured as described above will be described with reference to the flowchart shown in FIG. In step S1, the frame length counter N
0 (zero) is given as an initial value to each of c and the previous sample value Sx.

Next, in step S2, the analog signal output from the microphone 11 is
, And a frame division process is performed in step S3. Next, in step S4, it is determined whether or not exactly one frame division has been completed.

If it is determined in step S4 that the frame division processing has been completed, difference extraction processing is performed in step S5, and enlargement / reduction processing is subsequently performed in step S6. Then, in step S7, data to be output to the D / A converter 31 is stored, and the process proceeds to step S8. If it is determined in step S4 that the frame division processing has not been completed, the process immediately proceeds to step S8.

In step S8, one data stored in step S7 is output to the D / A converter 31.
Next, in step S9, it is determined whether or not a predetermined time τs has elapsed after output to the D / A converter 31.

If it is determined in step S9 that the predetermined time τs has not elapsed, the determination in step S9 is repeated. If it is determined in step S9 that the predetermined time τs has elapsed, the process returns to step S2, and the processes in steps S2 to S9 are repeated. Here, the predetermined time τs can be said to be a sampling (sampling) cycle of the audio signal processing device of the present invention.

Next, a detailed procedure of time frame division by the time frame dividing means 2 executed in step S3 shown in FIG. 3 will be described with reference to a flowchart shown in FIG. In step S11, the condition that the product of the new sample value S and the immediately preceding sample value Sx is equal to or smaller than 0 is satisfied, that is, whether at least one of them is zero or both signs are different. Make a judgment.

If it is determined in step S11 that the condition is not satisfied, the process proceeds to step S12. If it is determined that the condition is satisfied, the process proceeds to step S13. In step S12, each of the previous sample value Sx and the frame buffer FB [N] is set as a new sample value S, and the frame length counter Nc is incremented by one.
In step S13, it is determined whether or not the condition that the frame length counter Nc is 1 or more is satisfied. If it is determined that the condition is not satisfied, the process proceeds to step S12.

If it is determined in step S13 that the condition is satisfied, the process proceeds to step S14, where the frame length NN
To Nc and the frame buffers FB [0] to FB [Nc−
1] is stored as a time frame, the frame buffer FB [0] is set to a new sample value S, the frame length counter Nc is set to 1, and the immediately preceding sample value Sx is set to a new sample value S. And the process ends.

Here, if the processing is completed through step S14, it is determined in step S4 that one frame division is completed. If the processing is completed through step S12, frame division is completed in step S4. It is determined that it has not been completed.

Next, a detailed procedure of the difference calculation by the difference extracting means 5 executed in step S5 shown in FIG. 3 will be described with reference to a flowchart shown in FIG. The difference calculation is performed for two consecutive sample values included in the time frame.
The calculation is performed by calculating all the absolute values of the differences between the two sample values and calculating the average value.

In step S21, the absolute value D of the difference
And zero is assigned to each of the frame counter N and the process proceeds to step S22. In step S22,
It is determined whether or not the frame counter N satisfies a condition of being greater than or equal to a value obtained by subtracting 1 from the frame length NN.

If it is determined in step S22 that the condition is not satisfied, in step S23, a value obtained by subtracting the absolute value of the difference between the frame buffer FB [N + 1] and the frame buffer FB [N] from the absolute value D of the difference is used. At the same time, the frame counter N is incremented by one, and the process returns to step S22. If it is determined in step S22 that the condition is satisfied, in step S24, the absolute value D of the difference is set to a value obtained by dividing the absolute value D of the difference by a value obtained by subtracting 1 from the frame length. Save D and exit.

Next, a detailed procedure of signal enlargement / reduction by the signal enlargement / reduction means 6 executed in step S6 shown in FIG. 3 will be described with reference to a flowchart shown in FIG. In step S31, it is determined whether or not the condition that the absolute value D of the difference is larger than a predetermined threshold value TH is satisfied.

If it is determined in step S31 that the condition is not satisfied, the frame counter N is set to 0 in step S32, and in step S33, the condition that the frame counter N is greater than or equal to the frame length NN is satisfied. Judge whether or not.

If it is determined in step S33 that the condition is not satisfied, in step S34, the update frame buffer FBR [N] is set to 0, the frame counter N is incremented by 1, and the process returns to step S33. If it is determined in step S33 that the condition is satisfied, the process proceeds to step S43.

If it is determined in step S31 that the condition is satisfied, in step S35, the maximum value MAX of the absolute value of the sample value and the frame counter N are set to 0, and the process proceeds to step S36. In step S36, it is determined whether or not the frame counter N satisfies the condition of being greater than or equal to the frame length NN.

If it is determined in step S36 that the condition is not satisfied, in step S37, the absolute value of the frame buffer FB [N] is set to the maximum value M of the absolute values of the sample values.
It is determined whether or not the condition of being larger than AX is satisfied.

If it is determined in step S37 that the condition is satisfied, in step S38, the maximum value MAX of the absolute value of the sample value is set to the absolute value of the frame buffer FB [N], and then the flow proceeds to step S39. If it is determined in step S37 that the condition is not satisfied, step S3
Move to 9. In step S39, the frame counter N is incremented by one and the process returns to step S36.

If it is determined in step S36 that the condition is satisfied, in step S40, the frame counter N is set to 0, and the flow advances to step S41. Step S4
At 1, it is determined whether or not the frame counter N satisfies the condition of being greater than or equal to the frame length NN.

If it is determined in step S41 that the condition is not satisfied, in step S42, the product of the updated frame buffer FBR [N] and the constant value CONST is calculated as the maximum value of the absolute value of the sample value. MAX
And the frame counter N is incremented by one, and the process returns to step S41. If it is determined in step S41 that the condition is satisfied, in step S43, the update frame buffers FBR [0] to FBR are updated.
The data of [N-1] is stored as the updated time frame, and the processing ends.

In the signal enlarging / reducing means 6, when the average of the absolute value D of the difference exceeds a predetermined threshold value TH, the largest one of the sample values included in the time frame is set to a predetermined value. Multiply all sample values in the time frame by the same factor. As a result, the time frame signal containing the vowel or consonant is amplified so that its peak level becomes constant. In particular, low-level consonants are enlarged. Conversely, when the average of the absolute value D of the difference does not exceed the threshold value TH, the signal is reduced by multiplying all the sample values by 0 (or a small value) as a coefficient.

FIG. 7 shows the state of signal processing by the signal enlarging / reducing means 6. In FIG. 7, the waveform indicated by the solid line is D /
This is the output waveform of the A-converter 31, and the waveform indicated by the broken line is the output waveform of the microphone 11 before processing, that is, the output waveform of the microphone 11. FIG. 7 shows two adjacent sample values in each time frame.
This is a case where the average of the absolute values D of the differences between the two sample values exceeds a predetermined threshold value TH. Therefore, in FIG. 7, the amplitude in each time frame is amplified at the same ratio so that the maximum value of the amplitude in each of the time frames t1 and t2 becomes the predetermined maximum value M.

In the above-described audio signal processing apparatus, the speech sound is digitally sampled (sampled) to generate a sampled data sequence, which is divided by an appropriate time frame length, and the sampled data sequence is sampled within the time frame. Is calculated, the absolute value D of the difference between two adjacent sample values is calculated, and when the average value of the absolute values D of these differences exceeds a predetermined threshold value TH, the maximum value of the data sequence included in the time frame is calculated. Is multiplied by the same coefficient to all the data strings in the time frame so that に な る becomes a predetermined value, the data in the time frame including vowels or consonants is expanded so that its peak level is constant. . In particular, low-level consonants are enlarged more greatly.

If the average value of the absolute value D of the difference does not exceed the threshold value TH, the signal is reduced by multiplying all the sampled values by 0 as a coefficient, so that there is ambient noise. Even so, the level will be almost zero.

As an example, when the audio signal processing device is applied to a hearing aid, the output waveform (original waveform) of the microphone 11 when the sound of / ka / is input to the microphone 11 is shown in FIG. The output waveform of the / A converter 31 is shown in FIG.
Shown in As shown in FIG. 8A, in the sound of / ka /, a phoneme / k / which is a consonant part and a phoneme / a / which is a vowel part are output from the microphone 11 in series. Then, as shown in FIG. 8 (b), at the same time as the consonant (/ k /) is selectively amplified by the time frame dividing means 2 and the signal processing means 3, a conversation sound before and after the voice / ka / is produced. The area without the noise is reduced, and the level of the ambient noise is almost zero.

According to the present invention, consonants and ambient noise can be distinguished, and only consonants can be amplified more strongly than ambient noise when there is no conversational sound. For example, the difference between a plosive consonant and a fricative consonant is large because it has impact properties, and the frictional consonant contains many high frequency components even if the level is low.
On the other hand, the ambient noise is mainly composed of low frequency components, so that the difference becomes small and is distinguished from consonants.

At this time, since the vowel has a high level, the difference value is also large. Therefore, after the conversation sound is divided into time frames, only the time frame signal containing a large difference value is expanded to a certain level, and the other time frames are reduced in reverse, so that the conversation noise without surrounding noise is removed. It is possible to amplify only the sound and make the levels of the consonant and the vowel equal to make it easier for a hearing-impaired person to hear.

Further, the signal processing in the present invention is mainly for calculating only the difference value and does not perform the fast Fourier transform or the like, so that the length of the time frame can be freely changed. Therefore, the zero crossing point of the conversation sound signal can be set as the time frame boundary. As a result, even if the enlargement / reduction ratio of the signal is changed for each time frame, the discontinuity of the waveform does not occur and the sound quality becomes natural.

FIG. 9 shows an actual example of the effect of a hearing aid to which the audio signal processing device according to the present invention is applied. It is a judgment diagram at the time of hearing to 17 ears of a test subject who has it. At that time, the number of people who can hear each of the six phrases is shown in the table. For example, in the third cell from the left and the sixth cell from the top, 2 is written because only one signal was heard when no signal processing was performed, but four signals were heard when signal processing was performed. This indicates that there were two ears.

As a general tendency, those having a small number of pieces that can be heard without signal processing have a better number of hearings when signal processing is performed. Conversely, those that could be heard a lot without signal processing became worse. People who hear well without signal processing can be satisfied with conventional hearing aids, but those who do not have problems, and the present invention which can improve it is useful. In particular, it can be seen that more than half of the hearing-impaired persons whose unprocessed (in the case where the present invention is not used) the number of cognitive phrases is 0 to 1 and low is improved from 3 to 5 according to the present invention.

Further, according to the present invention, it is not necessary to execute a large number of multiplications unlike the FIR type digital filter, so that the amount of calculation is small, and therefore, the present invention can be realized by a small-scale signal processing device and its power consumption can be reduced. This enables a portable device such as a hearing aid, for example, to be smaller and have a longer battery life, and is highly useful.

In the above-described embodiment, the time frame is generated by dividing the time-series sample values by the boundary where the sign changes by the time frame dividing means 2. Alternatively, a plurality of boundaries where the sign changes may be included so as to be separated by the boundary where the sign changes.

When frames are generated at regular time intervals by the time frame dividing means 2, as shown in the flowchart of FIG. 10, in step S51, the frame buffer FB [N] is set to a new sample value S, and a frame length counter is set. Increment Nc by one. Next, in step S52, it is determined whether or not a condition that the frame length counter Nc is greater than or equal to the frame length NN is satisfied.

If it is determined in step S52 that the condition is satisfied, in step S53, the data of the frame buffers FB [0] to FB [N-1] is stored as a time frame, and the process ends. If it is determined in step S52 that the condition is not satisfied, the process ends.

When determining the coefficients for enlarging / reducing the signal of the time frame by the difference extracting means 5, all the differences between the sample values included in the time frame can be calculated and the average value thereof can be used. However, only the difference at the zero crossing point of the signal corresponding to the time frame boundary can be used as a representative. In each case, a good hearing effect can be obtained.

When only the difference at the zero-crossing point of the signal corresponding to the time frame boundary is used as a representative by the difference extracting means 5, as shown in the flowchart of FIG. The absolute value of the difference between [1] and the frame buffer FB [0] is stored, and the process ends.

In the above embodiment, when the average of the absolute value D of the difference does not exceed the threshold value TH, the signal is greatly reduced by multiplying all the sampled values by 0 (or a small value) as a coefficient. However, since this coefficient determines how much the consonant component is emphasized more than the other components including the vowel, it may be appropriately changed depending on the situation. For example, the value of this coefficient may be 1, or may be a value somewhat smaller than the coefficient by which the coefficient is multiplied when the threshold value is exceeded.

In short, if the absolute value D of the difference is large, the coefficient is increased, and if the absolute value D of the difference is relatively small, the coefficient is reduced accordingly. There is.

In the above-described embodiment, the case where the present invention is applied to a hearing aid is shown. However, the present invention relates to a communication device which requires a clear listening of a conversation sound, for example, a telephone and a radio. And so on.

[0057]

As described above, according to the first aspect of the present invention, consonants and ambient noise can be distinguished and only the level of consonants can be amplified. That is, since conversational sounds (especially consonants) and stationary ambient noises can be distinguished by simple signal processing and selectively amplified, conversational sounds can be easily heard, and this is particularly effective for hearing-impaired people.

According to the second aspect of the present invention, there is no discontinuous portion in the waveform after the signal processing, and an output sound with more natural sound quality can be obtained.

According to the third aspect of the present invention, in an environment where conversation sounds and ambient noise are mixed, only conversation sounds excluding ambient noise are selectively amplified, and the consonant and vowel levels are output at the same level. Can be easily heard by the listener.

According to the fourth aspect of the invention, it is possible to calculate an appropriate representative value of the difference between the sample values included in the time frame, that is, an index representative of the characteristic of the speech signal, and use this value to amplify the signal (expansion). / Reduction ratio), the distinction between the ambient noise and the conversation sound can be made clear.

According to the fifth aspect of the present invention, a reasonable representative value of the difference between the sample values included in the time frame can be calculated in the same manner as in the fourth aspect of the present invention. be able to.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a conceptual configuration of an audio signal processing device according to the present invention.

FIG. 2 is a configuration diagram of an audio signal processing device.

FIG. 3 is a flowchart showing a processing procedure of the audio signal processing device.

FIG. 4 is a flowchart showing a procedure of time frame division.

FIG. 5 is a flowchart showing a procedure for calculating a difference.

FIG. 6 is a flowchart showing a procedure of signal enlargement / reduction.

FIG. 7 is a waveform chart showing a signal amplified by the audio signal processing device.

FIG. 8 is a diagram showing an input waveform (a) and a processed waveform (b).

FIG. 9 is a determination diagram showing an actual example of the effect of a hearing aid to which the audio signal processing device according to the present invention is applied;

FIG. 10 is a flowchart showing the procedure of another embodiment of time frame division.

FIG. 11 is a flowchart showing a procedure of another embodiment of difference extraction.

[Explanation of symbols]

1 ... external sound sampling means, 2 ... time frame dividing means, 3
... Signal processing means, 4 ... Output means, 5 ... Difference extraction means, 6
... Signal enlargement / reduction means, 7 ... Operation unit, 11 ... Microphone, 12 ... A / D converter, 21 ... CPU, 22 ...
ROM, 23 ... RAM, 31 ... D / A converter, 32 ... Earphone.

Claims (5)

[Claims] 1. A sampling means for sampling a speech signal at regular intervals to generate time-series sample values, and a time frame of sample values by dividing the sample values into a plurality of successive values. Time frame dividing means for generating, signal processing means for performing signal processing on the time frame and updating the sample value included in the time frame, and output for converting the updated sample value into an audio signal and outputting the signal Means, the signal processing means, a difference extraction means for calculating the absolute value of the difference between two consecutive sample values, and a coefficient determined according to the calculated absolute value of the difference to the sample value An audio signal processing device comprising a multiplying signal enlarging / reducing means. 2. The audio signal processing apparatus according to claim 1, wherein the time frame dividing means divides the sample value at a boundary where the sign changes so as to correspond to a zero crossing point of the speech signal. 3. The signal enlarging / reducing means, when the absolute value of the difference between the sample values exceeds a predetermined threshold value, sets a maximum absolute value of the sample value included in the time frame to a constant value. The coefficient is determined so that the absolute value of the difference between the sample values does not exceed the threshold value, and the maximum value of the absolute value of the sample value included in the time frame becomes smaller than a certain value. The audio signal processing device according to claim 1, wherein the coefficient is determined as follows. 4. The method according to claim 1, wherein the difference extracting means calculates all the absolute values of two consecutive differences with respect to all of the sample values included in the time frame, and further calculates an average value thereof. 4. The audio signal processing device according to 1, 2, or 3. 5. The difference extracting means calculates an absolute value of a difference between the first sample value and / or the last sample value included in the time frame and successive sample values. The audio signal processing device according to claim 1, 2, 3, or 4.