JPH0698398A - Non-voice section detecting/expanding device/method - Google Patents

Abstract

PURPOSE:To prevent a non-voice section detecting/expanding device/method which controls the expanded variable of the non-voice section in response to the sound power. CONSTITUTION:The input voice signals are stored in a memory 111 and also in an envelope curve detecting means 12. The detected envelope curve information is stored in a memory 112 and then in a maximum/minimum value detecting means 13. Then the maximum and minimum values are stored in a memory 113. The voice are outputted being expanded between those memories.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to audio signal processing for assisting a deteriorated auditory characteristic used in a hearing aid or the like,
In particular, the present invention relates to an apparatus for detecting and expanding a silent section of a voice for detecting and expanding a silent section in a voice, a method for detecting and expanding a silent section of a voice, and a device using the same.

[0002]

2. Description of the Related Art In a hearing aid that assists the hearing of a hearing-impaired person, an analog-type hearing aid that processes an amplitude and a frequency characteristic of a voice using an analog circuit has been mainly used. On the other hand, in recent years, research and development of digital hearing aids applying digital signal processing have been actively conducted. Regarding this research and development trend, for example, the Acoustical Society of Japan (19)
1989, Vol. 45, No. 7, P549-P555) "Recent trends in hearing aids" and the like. Audio signal processing in this digital hearing aid is performed by digital signal processing using a digital signal processor (hereinafter abbreviated as DSP). The processing contents are described by the program. Therefore, compared to conventional hearing aids, digital hearing aids can change the contents of voice processing by changing the program on the memory, and adjustments are made to maximize the intelligibility of voices for individual hearing impaired persons. Can be done more easily. Further, in the sense that the digital hearing aid replaces the analog hearing aid, it is premised that all necessary processing is completed in real time, that is, within a time delay that is not felt by the user.

Audio signal processing used in a digital hearing aid includes processing for compensating for the deterioration of the frequency resolution, time resolution, spectral discrimination ability, sound image synthesis ability, etc. of a hearing-impaired person. For these processes, for example, the Journal of Acoustical Society of Japan (198
7:43, No. 5, P 356 to P 361) “Digital hearing aid with audio feature extraction function” and the like. Among them, as one of the processes for compensating for the deterioration of the auditory function, especially the time resolution, there is a method of extending a silent section in a voice. The method of extending this silent section is
By inserting a small silence interval between the consonant part and the vowel part of the speech, or by extending the length of the silence interval between words, it is intended to prevent the above-mentioned continuous masking. At the same time as it reduces masking, it also has the effect of compensating for the slowdown in voice understanding.

In the case of a sensorineural hearing-impaired person, which is common in the elderly, in order to process the voice and improve its intelligibility,
It is desired that not only the frequency characteristics of the voice are processed to assist the frequency resolution of the user, but also the processing for compensating for the deterioration of the time resolution is simultaneously performed. As a concrete example of the deterioration of the time resolution, there is a phenomenon of "continuous masking" in which a strong voice component emitted immediately before makes it impossible to hear a weak voice component immediately after. Due to this phenomenon, a hearing-impaired person often falls into a situation where "sound can be heard but cannot be heard as words". This phenomenon has been reported in detail in, for example, Technical Report of IEICE (SP90-97) "Effect of continuous masking on voiced / unvoiced perception of plosives".

[0005]

In the conventional research on the technology relating to the detection and extension of the silent section for the purpose of hearing aid, in a laboratory level experiment, a speech waveform is displayed on a CRT screen and the waveform shape is visually observed. And S / N
A method of processing with a simple threshold using only audio material with a high ratio has been adopted.

In order to perform the above-described expansion processing of the silent section, it is necessary to detect the silent section in the voice waveform in the time domain. Since the voice waveform is not stationary due to the influence of background noise, etc., in the detection of the silent section, the silence that is not easily affected even if the voice level changes due to the noise component according to the state of the voice waveform at that time. There is a problem in that an algorithm and parameters for detecting the section must be selected appropriately.

However, the above-mentioned conventional technique merely provides a simple method for detecting a silent interval, and does not dynamically expand the silent interval in real time according to an input voice signal. For this reason, it is not a means for solving the practical problems that may occur during use as described above.

An object of the present invention is to solve the above problems and to provide an apparatus for detecting and expanding a silent section of a voice and a method for detecting and expanding a silent section of a voice for detecting and expanding a silent section in a voice in order to assist a deteriorated auditory characteristic. To provide. More specifically, the present invention provides a silent interval detection / expansion device and method that can control the expansion amount of a silent interval according to the power of voice and is less susceptible to noise components, and a device using the same. To do.

[0009]

In order to achieve the above object, a silent section is detected from an input voice signal and the silent section is extended. At this time, the degree of expansion of the silent section is controlled according to the power of the voice in order to suppress continuous masking, instead of merely mechanically inserting the silent section for a fixed time. In the detection of silent sections, it is less susceptible to the effects of background noise,
The voice power is detected and used as a parameter for expanding the silent section, and the hearing is assisted by expanding the silent section.

That is, in a soundless section detecting and expanding device for aurally assisting a voice which inputs a sound, performs a sound waveform processing, and outputs a sound, a silent section detecting means for detecting a silent section in an input sound signal and the silent section. A silent interval signal detected by the interval detecting means is repeatedly added to the silent interval of the input voice signal to expand the silent interval, and the silent interval expanding means is included. The silent section is extended in proportion to the power of the input audio signal. In addition, the silent section expansion means repeatedly adds a part at any time point within the silent section including the start point and the end point of the silent section to the silent section.

Further, in a method for detecting and expanding a silent section of a voice for assisting hearing, which inputs a voice, performs a voice waveform processing, and outputs a voice, a silent section in an input voice signal is detected and a signal in the silent section is input. This is a silent interval detection / expansion method that expands the silent interval by repeatedly adding it to the silent interval of the audio signal, and makes the amount of signal repetition in the silent interval proportional to the power of the input audio signal immediately before the silent interval, A part of the signal of the silent section is repeated and added to the silent section at any point in the silent section including the start point and the end point.

More specifically, the first means for achieving the above object is to record the voice waveform to be subjected to the expansion processing of the silent section in the memory, and detect the envelope of the voice waveform simultaneously with the recording. Means for detecting the maximum value and the minimum value of the envelope, and means for calculating the threshold value for the detection of the silent section from the maximum value and the minimum value, to reproduce the recorded voice. First, the part where the envelope of the voice is continuously below the threshold is detected as a silent section.

A second means for achieving the above object is a means for detecting an envelope curve of a voice waveform to be subjected to a silent zone expansion process, and a means for calculating a differential value of the envelope waveform. The time required for the differential value to change from negative to positive is determined as the silent section, and the amount of expansion of the silent section is determined in proportion to the peak value of the differential value.

Further, a third means for achieving the above object is a means for detecting an envelope of a voice waveform to be subjected to a silent zone expansion process, and a means for delaying the envelope waveform for a predetermined time. OF that detects rising and falling edges of the envelope from the difference between the integral value of the envelope and the integrated value of the delayed envelope
An F neuron circuit is provided, the OFF neuron circuit discriminates the time from the detection of the falling edge of the envelope to the detection of the rising edge as a silent section, and the OFF neuron when detecting the rising and falling edges of the envelope. Determines the amount by which the silent interval is extended in proportion to the output of the circuit.

[0015]

In the first means, the envelope of the voice waveform is detected by the means for detecting the envelope of the voice waveform, and the input voice and the envelope are simultaneously recorded in a recording medium such as a memory. Envelope detection at the time of reproduction can be omitted, which is convenient for expanding the silent section at the time of reproduction in real time. Further, by determining the threshold value for the detection of the silent section from the maximum value and the minimum value of the recorded envelope information, it becomes possible to set the threshold value according to the level fluctuation of the speech to be processed, Even if the level of the voice changes due to the component, it is possible to detect the silent section that is hardly affected by the change.

In the second means, since the rising and falling edges of the voice are detected based on the differential value of the envelope, the silent section can be detected even when a stationary noise component is included in the voice. It becomes possible to do it. Further, since the amount of expansion of the silent section is determined in proportion to the peak value of the differential value, the silent section after the steep falling is expanded long, and the silent section after the gentle falling is not expanded very much. As a result, the silent section after the voice having a strong power where continuous masking easily occurs is further expanded, and it is possible to realize the expansion of the silent section according to the voice power, which is suitable for preventing continuous masking. It will be possible.

In the third means, the OFF neuron circuit takes the difference between the envelope and the delayed envelope to detect the rising and falling edges of the input voice. Therefore, even if the voice includes a stationary noise component, the influence can be reduced by taking the difference as the difference. Also, since the output of the OFF neuron circuit is proportional to the value of the envelope, by determining the expansion amount of the silent section in proportion to the output of this OFF neuron circuit, it is possible to expand the silent section according to the power of the voice. Becomes

[0018]

EXAMPLES The present invention will be described in detail below with reference to examples. FIG. 1 is a block diagram of a silent interval expansion means 1 using envelope threshold processing, which is a first embodiment of the present invention. Silent section decompression means 1 using this envelope threshold processing
The outline of is as follows.

The input voice signal is an A / D converter (not shown)
An envelope detecting means 1 for detecting the envelope of the input voice signal at the same time that the input voice signal converted into a digital signal by the
Entered in 2. The detected envelope information is stored in the envelope information storage means 112 of the memory 11, and further the maximum value minimum value detection means 13 for detecting the maximum value and the minimum value of the envelope.
The maximum value and the minimum value are stored in the maximum value / minimum value storage means 113 of the memory 11. When reproducing the voice stored in the memory 11, the voice information storage means 111,
Envelope information storage means 112 and maximum / minimum value storage means 1
The content of No. 13 is input to the silent section expansion means 14, and the silent section in the voice information Data is set to the envelope information Env. And the maximum value Max.
Detect using the minimum value Min., Expand the silent section,
Outputs the voice in which the silent section is expanded.

More specifically, the memory 11 for storing information is divided into three parts inside. The input voice signal is stored in the voice information storage means 111 and simultaneously input to the envelope detection means 12. The envelope detecting means stores the detected envelope information Env. In the envelope information storing means 112 of the memory 11. Further, the envelope information is input to the maximum value / minimum value detection means 13 for detecting the maximum value and the minimum value of the envelope, and the maximum value Max. And the minimum value Min. Of the detected envelope are the maximum and minimum values of the memory 11. It is stored in the value storage means 113. The detection of the envelope is realized by, for example, performing time averaging one after another on the average value of a certain section of the input voice data. As a result, the memory 11 stores the voice information input within a certain period of time, the envelope information temporally corresponding thereto, and the maximum and minimum values of the envelope information within the certain period of time. When reproducing the voice stored in the memory 11, the voice information Data stored in the voice information storage unit 111 of the memory 11 and the envelope information Env. Stored in the envelope information storage unit 112 of the memory 11. , The maximum value of the envelope stored in the maximum value minimum value storage means 113 of the memory 11
The Max. Minimum value Min. Is input to the silent section expansion means 14. The silent section expansion means 14 is a voice information storage means 11.
The silent section in the voice information given from 1 is detected using the envelope information given from the envelope storing means 112 and the maximum / minimum values of the envelope given from the maximum / minimum value storing means 113, and then detected. The silent section is expanded, and the voice with the expanded silent section is output.

The details of the silent section expansion means 14 are shown in FIG. In the silent section expansion means 14, the threshold value setting means 141 first sets the threshold value T from the maximum value Max. Minimum value Min. Of the envelope given by the maximum value minimum value storage means 113. The value of this threshold value T is the maximum value minimum value storage means 113.
It is set to any value between the maximum value Max. And the minimum value Min. Of the envelope given by. Next, the silent section detecting means 14
2 compares the envelope information Env. Given from the envelope information storage means 112 with the threshold value T, and detects a period in which the value of the envelope information becomes smaller than the threshold value T as a silent section. (A period in which the value of the envelope information is continuously lower than a preset threshold value is obtained, and if the period is longer than the preset shortest silent period, it is detected as a silent period.) Next, waveform processing The means 143 sets the period corresponding to the period detected as the silent section by the silent section detecting means 142 on the voice information given from the voice information storage means 111. When outputting the voice information, the voice information other than the silent section is output as it is, and the voice information corresponding to the silent section is repeatedly output. Further, the repetition amount is set to a value of the envelope of the voiced section immediately before the repeated silent section, or a value proportional to the maximum value of the envelope given from the maximum / minimum value storage unit 113.
However, this repetition amount is not limited to an integer, and a real number such as 1.2 times or 3.4 times can be selected. As a result, it is possible to expand the silent section according to the size of the audio signal.

FIG. 3 shows a state of the expansion processing of the silent section by using an example of a voice waveform. In this example,
The periods when the envelope amplitude is smaller than the threshold value T are t1 and t
It is located at two locations of 2 and is expanded a times and b times respectively and output. The values of a and b are determined using the envelope amplitude value immediately before each silent section. Since the threshold value T is always set to a value between the maximum value and the minimum value of the envelope recorded in the memory, it is always possible to detect a period with a small envelope amplitude as a silent section. According to the present embodiment, the envelope information and the parameter for determining the threshold value are detected and recorded when the voice is recorded, so that the amount of signal processing at the time of reproduction becomes extremely small. For this reason, other processing that requires a long processing time during reproduction, such as audio frequency characteristic processing,
When performing simultaneously in real time, the effect is large in terms of reduction of signal processing amount. In this embodiment, the case where the input voice signal is digitized and the input voice signal is processed has been described. However, the input voice signal is not digitized, but the input voice signal remains as an analog signal. The detected maximum value and minimum value are digitized and stored in the memory, the input voice signal and the detected envelope information are stored in a storage medium in an analog manner, and a silent section is detected by analog processing to process the voice waveform. It goes without saying that the silent section can be extended. Next, FIG. 4 shows a block diagram of the silent section expansion means 2 using the differential value of the envelope, which is the second embodiment of the present invention.

The input voice signal is an A / D converter (not shown)
The input voice signal converted into a digital signal by the input signal is input to the envelope detecting means 22 and simultaneously to a memory (not shown) provided in the delay circuit 21. The envelope information Env. Detected by the envelope detecting means 22 is input to the differentiating circuit 23. Differentiating circuit 23
Is a circuit that calculates and outputs the differential value of the envelope information that is continuously input, and the output changes according to the rising and falling edges of the envelope. An example of this differentiating circuit is shown in FIG.
There is a method of obtaining the differential value by convolving the window function W (τ) with the envelope data as shown in FIG. This can be expressed as a mathematical expression as follows.

[0024]

[Equation 1]

On the other hand, the delay circuit 21 delays the input voice signal by the time required for the envelope detection and the calculation of its differential value, and sends the input voice signal to the silent section expansion means 24. Next, the output D-out of the delay circuit and the output Diff. Of the differentiating circuit are input to the silent section expansion means 24, and the silent section is detected and the silent section is expanded.

FIG. 6 shows details of the silent section expansion means 24. Of these, the silent section detecting means 241 detects a period in which the output Diff. Of the differentiating circuit 23 changes from a negative peak value to a positive peak value as a silent section. (Calculate the period from the negative peak value to the positive peak value,
When the period is longer than the preset shortest silent section, it is detected as a silent section. The waveform processing means 242 outputs the output of the delay circuit 21 based on the detection result.
Expands the silent section in the D-out audio signal. The waveform processing performed here is performed by the same method as that of the first embodiment described above.

FIG. 7 shows an example of the speech waveform of the expansion processing of the silent section according to the second embodiment. The output amplitude of the differentiating circuit has a positive peak at the transition part from the silent section to the sound section in the input voice, and a negative peak at the transition section from the sound section to the silent section. The peak interval from negative to positive is determined as a silent interval, and the expansion amount of the silent interval, that is, the repetition amount when outputting the silent interval is determined in proportion to the absolute value of the negative peak that occurs immediately before the silent interval. . In this example, there are two locations, t1 and t2, and they are output after being expanded a times and b times, respectively. Note that here, the window function W (τ)
Although the shape of has been described using a straight line with a positive slope, even when a window function with an inverse slope is used, the period during which the output of the differentiator circuit changes from positive to negative is regarded as a silent interval. It goes without saying that the same effect as described above can be obtained by the detection. Further, it is obvious that the window function W (τ) also has a non-linear shape as shown in FIG. 5B, and the same effect can be obtained by performing weighted addition.

According to the present embodiment, since the detection of the silent section is detected by the change of the envelope, even if the stationary noise is superposed on the voice, only the envelope signal becomes large, and the silence signal is generated. It has little effect on the detection of sections. In this embodiment, the case where the input voice signal is digitized and the input voice signal is processed has been described. However, the input voice signal is not digitized, but the input voice signal remains as an analog signal. It goes without saying that it is possible to detect and give an analog delay to the input voice signal, differentiate the envelope information by analog processing, detect the silent section, process the voice waveform, and extend the silent section. At this time, the mathematical expression for obtaining the differential value by convolving the window function with the envelope data to obtain the differential value may be analog processing by replacing the addition symbol Σ with the integration symbol ∫ in the above (Equation 1).

Next, an OF, which is a third embodiment of the present invention, will be described.
FIG. 8 shows a block diagram of the silent section expansion means 3 using the F neuron circuit.

The input audio signal is an A / D converter (not shown)
The input voice signal converted into a digital signal by the input signal is input to the envelope detecting means 32 of the silent section detecting means and to the delay circuit (2) 31 at the same time as in the second embodiment. Provided memory (not shown)
Is entered. The envelope information detected by the envelope detecting means 32 is stored in a memory (not shown) provided in the delay circuit (1) 35.
Entered in. The OFF neuron circuit 33 configured by a digital circuit includes the envelope information Env. Detected by the envelope detecting means 32 and the envelope information D-env. Delayed by the delay circuit 35.
A signal corresponding to the rising and falling edges of the envelope
Output N-out. On the other hand, the delay circuit (2) 31 is
The input voice is delayed by the time required to obtain the output of the F neuron circuit 33. And this delayed input voice
The D-out is input to the silent section expansion means 34. The silent section expansion means 34 detects and expands a silent section in the delayed input voice D-out by using the output N-out of the OFF neuron circuit and outputs it. The OFF neuron circuit 3
3 detects the rising and falling edges of the envelope from the envelope information Env. And the delayed envelope information D-env., That is, the transition period from the voiced section to the silent section in the voice. The neuron group called OFF neuron, which is widely present in the visual and auditory sense systems of the living body, is a neural tissue that specifically reacts only when an input signal that has been applied turns OFF. Plays an important role in.
The OFF neuron circuit is a circuit that partially simulates the operation of the OFF neuron and detects the change point of the input signal. As shown in FIG. 9, for example, the OFF neuron circuit weights the envelope signal Env. That is input momentarily and the delayed signal D-env. It can be realized. Here, the weight Wi applied to the envelope information has a negative value, and the delayed weight We applied to the envelope information has a positive value. The absolute values of these two weights are made equal. As a result, the OFF neuron circuit outputs a negative value during the rising period of the envelope and a positive value during the falling period. Further, the magnitude of the output has a value corresponding to the magnitude of the envelope of the input voice. Further OF
Since the two input signals of the F neuron circuit are exactly the same signals differing only in time delay, even if the input voice contains stationary noise, this effect is canceled by the weighted addition operation. Output does not occur. On the other hand, the silent section expansion means 34 detects a period from when the output of the OFF neuron circuit outputs a positive value until the next negative value is output as a silent section (outputs a positive value. Then, a period for outputting the next negative value is obtained, and when the period is longer than a preset shortest silent period, it is detected as a silent period.), And the input delayed by the delay circuit (2) 31. The corresponding silent interval portion in the signal is repeatedly output to realize the expansion of the silent interval. Here, by setting the repetition amount of the silent section to be proportional to the output of the OFF neuron circuit used when detecting the silent section, the expansion of the silent section according to the volume of the input voice can be realized.

FIG. 10 shows an example of a voice waveform showing the process of expanding the silent section. The weighted addition result of the envelope of the input signal and the delayed envelope is the OFF neuron circuit output. In this example, since We has a positive value and Wi has a negative value, the OFF neuron circuit output is negative at the transition part from the silent section to the sound section in the input speech, and the transition part from the sound section to the silence section. Produces a positive peak at. The interval between the positive and negative peaks is determined as a silent section, and the expansion amount of the silent section, that is, the repetition amount when outputting the silent section is determined in proportion to the absolute value of the positive peak generated immediately before the silent section. In this example, there are two locations, t1 and t2, each of which is a
It is output after being doubled and doubled.

By reversing the positive and negative of We and Wi of the OFF neuron circuit, the positive and negative of the output of the OFF neuron circuit is reversed, and the OFF neuron circuit output changes from a negative value to a positive value next. Needless to say, even if is detected as a silent section, the same effect can be obtained. In this embodiment, the case where the input voice signal is digitized and the input voice signal is processed has been described. However, the input voice signal is not digitized, but the input voice signal remains as an analog signal. It goes without saying that it is possible to detect and provide an analog delay to the envelope information and the input voice signal, and to utilize the output of the OFF neuron circuit composed of an analog circuit to detect the silent section and extend the silent section. .

By the way, in the above embodiment, the silent section is expanded by repeatedly outputting the part detected as the silent section. Often included in part. In such a case, a short voice is repeatedly output and becomes new noise. In order to avoid this, it is possible to use a method of not repeating all the portions detected as the silent section but repeating the central portion excluding a part of both ends of the silent section. On the other hand, when the input voice information includes noise other than voice, the silent section detected by the envelope threshold processing often also includes noise. In such a case, if the above-mentioned silent section is expanded, noise will be repeatedly output, and new noise components that are periodically heard will be generated. In order to avoid this, a method of lowering the signal level and repeating it can be used when repeating the silent section.

On the other hand, in the above description, one audio signal has been described as an object, but humans usually listen to the sound using both left and right ears, and even in a device such as a hearing aid, it is used for both left and right ears. It is desirable to perform 2-channel processing.
In such a case, it is conceivable that the silent section is detected in one of the channels and the expansion processing is performed in two channels. In addition, in selecting a channel for detecting a silent section, there is a method in which a channel on the side of the user's effective ear is used as a detection channel, or a method in which a larger envelope signal is used as a detection channel. . Further, a method of calculating the average value of the outputs of both channels and detecting the silent section by using the envelope of the average value signal is also possible.

By the way, the silent section detecting / expanding means of the present invention can be applied to a communication device such as a telephone. The above-described detection and expansion processing of the silent section of the voice for the input audio signal to the hearing aid is added to the communication device of the communication device such as the telephone to expand the silent section of the voice output from the handset and decay. Can assist hearing. When the present invention is applied to the receiving side through the telephone line, a time lag with the transmitting side becomes a problem, but the transmitting side is informed in advance that a time delay will occur, and an appropriate pause is made during the conversation. There is no problem in practical use.

FIG. 11 is a diagram showing a configuration in which the first, second, and third silent section detecting / expanding means of the present invention are applied to a telephone. In the figure, a receiver circuit 41 of the telephone 4 is a circuit for extracting a voice signal from a signal sent through a telephone line, and a transmitter circuit 42, on the contrary, transmits a voice signal obtained by a microphone 432 in a handset 43 to a telephone. It is a circuit that converts the signal for the line and sends it to the telephone line. The silent section expansion means 40 is a silent section expansion means for voice (see FIG. 1, FIG. 4, FIG. 8 or the like) using at least one of the first to third means. When the telephone line is the analog line system, the voice signal is not digitally encoded, and therefore the analog voice signal extracted through the receiving circuit is once converted into a digital signal by using the A / D converter 48. The speaker 4 in the handset 43 is converted to an analog signal by the D / A converter 49 by performing the silent section expansion process described above.
Output from 31. In the case of the digital line system, the signal transmitted is already digitally encoded, and the decoding process which is carried out by an ordinary digital telephone is performed in the receiving circuit 41 to be converted into a digital voice signal. After that, without using the A / D converter 48, the above-mentioned silent interval expansion processing is performed by the silent interval expansion means, and the D / A converter 4
Needless to say, the same effect can be obtained by converting into an analog signal and outputting the analog signal.

On the other hand, the telephone according to the present embodiment has the extension amount changing means 44 for changing the extension amount by the user. The user changes the extension amount while listening to the voice during the call by means such as turning the volume, and performs the call at the extension amount that is the most audible. Further, the present embodiment has a parameter storage means 45 for storing the signal processing parameter indicating the expansion amount, and stores, for example, the parameter used in the call at the end of the call. When the user wants to make a call under the same conditions as those used in the past when making a new call, the user selects the parameters stored in the parameter storage means 45 by the parameter selection means 46. One of the sets is selected, and this parameter is set to the silent section expansion means 40.
Transfer to.

Further, the parameter selecting means 46 of the telephone of this embodiment is provided with a frequency detecting means 47. The frequency detection means 47 has a function of detecting the parameter set used most frequently among the plurality of parameter sets stored in the parameter storage means 45, and the parameter set most frequently used. Is used as an initial parameter set that is initially set when a call is started by this telephone. When the user wants to make a call with another parameter set, the parameter can be selected again by the parameter selecting means 46.

The expansion amount changing means 44, the parameter storing means 45, the parameter selecting means 46, etc. are not provided,
It is also possible to set a default value in the silent section decompressing means 40 in advance, and have a simple configuration in which the selection / non-selection of the silent section expansion can be simply designated by an external switch operation or the like.

In the above description, the silent section expanding means is used on the receiving side, but conversely, the silent section detecting and expanding means of the present invention is inserted immediately after the microphone on the transmitting side so that the speaker side can reproduce the voice. If processing is performed, it is possible to assist the hearing side's weakened hearing without imposing any burden on the receiving side. Further, in this case, as compared with the case where the present invention is used on the receiving side, the S / N of the voice to be processed is improved, so that there is an advantage that the silent section can be accurately detected and expanded. Further, the present invention can be applied to a so-called answering machine, and can also be used to assist listening of recorded voice.

Further, the silent section expansion means in the present invention can be applied to consumer equipment which outputs an audio signal, for example, equipment having audio output such as radio and television. FIG. 12 is a diagram showing an embodiment in which the first to third silent section detecting / expanding means of the present invention are used in a television receiver. The television signal included in the broadcast radio wave is extracted by the television radio wave receiving circuit 51, and the extracted television signal is separated into a video signal and an audio signal by the video / audio signal separating means 52. Of these, the video signal is processed by the image signal processing circuit 54 and then displayed on the display 57.

On the other hand, the separated audio signal is converted into an analog signal in the audio band by the audio signal processing circuit 53, digitized by the A / D converter 48, and passed to the silent section expansion means 40. This voice feature enhancement means 40
Is a silent section extending means (see FIG. 1, FIG. 4, or FIG. 8 etc.) using at least one of the first to third silent section detecting and extending means as in the embodiment of the telephone of FIG. The silent section of the voice described above is detected and expanded. The digital audio signal that has undergone the silent section expansion is D / A
After being converted into an analog signal by the converter 49, it is output by the speaker 56.

In the television receiver of this embodiment, the parameter storage means 45 and the parameter selection means 4 are included.
6 and an expansion amount changing means 44 are provided, but the functions of these means are the same as those of the embodiment of the telephone in FIG. Note that, also in the present embodiment, these means are not provided, a default value is set in advance in the silent section expansion means 40, and a switch operation from the outside, etc.
It is also possible to adopt a simple configuration in which selection / non-selection of emphasis is simply designated.

FIG. 13 is a diagram showing an embodiment in which the silent section detecting / expanding means of the present invention is used in a radio receiver. The radio signal included in the broadcast wave is extracted by the radio wave reception circuit 61, converted into an analog signal in the voice band by the voice signal processing circuit 62, digitized by the A / D converter 48, and the silent section expansion means 40. Passed to. As in the embodiment of the telephone of FIG. 11, the silent section expansion means 40 is a silent section expansion means (FIG. 1, FIG. 4, or FIG. 4) using at least one of the first to third silent section detection expanding means. 8 etc.), the silent section of the voice described above is detected and expanded. The digital audio signal subjected to the silent section expansion is converted into an analog signal by the D / A converter 49, and then output by the speaker 64.

The radio receiver of this embodiment is also provided with a parameter storage means 45, a parameter selection means 46, and an expansion amount change means 44. The functions of these means are those of the telephone set of FIG. It is similar to the embodiment. In addition,
Also in this embodiment, without providing these means, a default value is set in advance in the silent section expansion means 40, and selection / non-selection of emphasis can be simply designated by a switch operation from the outside. It is also possible to have a simple configuration.

The application of the present invention is not limited to the telephone set, the television receiver, and the radio receiver shown in FIGS. 11 to 13, and can be applied to a wide range of voice utilizing devices. For example, tape recorders, VTRs (video tape recorders), CDs (compact discs), DCCs (digital compact cassettes), MDs (minidiscs), and other audio recording devices,
It can be applied to voice output devices connected to WS (workstation), PC (personal computer), etc., WP (word processor) equipped with voice reading function and electronic mail, as well as devices and systems in industrial application fields. . In addition, in an educational site of a hearing-impaired child or the like, it can be applied to voice output of a microphone to assist the hearing of a plurality of hearing-impaired persons. Furthermore, it is also obvious that the silent section detection method of the present invention can also be used for segmentation of a voice signal in automatic voice recognition processing.

The silent section detecting and expanding means of the present invention can be easily realized by using a general-purpose DSP, but may be realized by software by using dedicated hardware or a general-purpose microcomputer. .

[0048]

According to the present invention, it is possible to detect a silent section without being affected by noise even if there is a change in voice level due to a noise component, and the silent section can be detected according to the magnitude of the power of the voice signal. It is possible to control the extension amount of. That is, the silent section can be dynamically detected in real time according to the input voice signal and the silent section can be extended. By applying the device and method of the present invention to a device such as a hearing aid or the like that assists a deteriorated hearing, or a consumer device that outputs a voice signal, the time resolution of a hearing-impaired person, which has not been considered so far, is reduced. It is possible to compensate for the decrease in voice intelligibility due to continuous masking and improve the voice intelligibility for the hearing impaired.

[Brief description of drawings]

FIG. 1 is a block diagram showing a first embodiment of the present invention.

FIG. 2 is a block diagram showing details of a silent section expansion means in FIG.

FIG. 3 is a diagram illustrating a process of expanding a silent section using envelope threshold processing.

FIG. 4 is a block diagram showing a second embodiment of the present invention.

FIG. 5 is a diagram showing an example of a window function used to obtain a differential value.

FIG. 6 is a block diagram showing details of a decompression means for a silent section in FIG.

FIG. 7 is a diagram showing a process of expanding a silent section using a differential value of an envelope.

FIG. 8 is a block diagram showing a third embodiment of the present invention.

9 is a diagram illustrating details of an OFF neuron circuit in FIG.

FIG. 10 is a diagram showing a decompression process of a silent section using an OFF neuron circuit.

FIG. 11 is a block diagram of a telephone using the silent interval expansion means of the present invention.

FIG. 12 is a block diagram of a television receiver using the silent section expansion means of the present invention.

FIG. 13 is a configuration diagram of a radio receiver using the silent section expansion means of the present invention.

[Explanation of symbols]

 1 ... Silent section expansion means using envelope threshold processing 11 ... Memory 111 ... Voice information storage means 112 ... Envelope information storage means 113 ... Maximum value minimum value storage means 12 ... Envelope detection means 13 ... Maximum value minimum value detection Means 14 ... Silent section expanding means 141 ... Threshold setting means 142 ... Silent section detecting means 143 ... Waveform processing means 2 ... Silent section expanding means 21 using differential value of envelope 21 ... Delay circuit 22 ... Envelope detecting means 23 ... Differentiating circuit 24 ... Silent section expanding means 241 ... Silent section detecting means 242 ... Waveform processing means 3 ... Silent section expanding means using OFF neuron circuit 31 ... Delay circuit (2) 32 ... Envelope detecting means 33 ... OFF neuron circuit 34 ... Silent Interval expansion means 35 ... Delay circuit (1) Wi ... Negative value weighting coefficient We ... Positive value weighting coefficient, 4 ... Silent interval expansion means Telephone set 41 ... Receiving circuit 42 ... Sending circuit 43 ... Handset 44 ... Expansion amount changing means 45 ... Parameter storing means 46 ... Parameter selecting means 47 ... Frequency detecting means 5 ... Television receiver 51 having silent section expanding means 51 ... Television radio wave receiving circuit 52 ... Video / audio signal separating means 53 ... Audio signal processing circuit 54 ... Image signal processing circuit 6 ... Radio receiver 61 having silent section extending means 61 ... Radio electric wave receiving circuit 62 ... Audio signal processing circuit.

Claims (37)

[Claims] 1. A device for detecting and expanding a silent section of a voice for assisting hearing, comprising: a silent section detecting unit for detecting a silent section in an input voice signal; and a silent section signal detected by the silent section detecting unit. A silent interval detecting / expanding device for audio, comprising: a silent interval expanding means for expanding in proportion to the power of the input audio signal immediately before the silent interval. 2. An envelope detecting means for detecting an envelope from an input voice signal converted into a digital signal, a means for detecting a maximum value and a minimum value of the envelope, and the input voice according to claim 1. A voice information of a signal, information of the envelope, a memory for storing at least the maximum value and the minimum value, and the silent section detecting means is a threshold value set to any value between the maximum value and the minimum value. An apparatus for detecting and expanding a silent section of a voice, which detects a period in which an envelope becomes smaller than that of a silent section. 3. The envelope detecting means according to claim 2, wherein the envelope curve of the input voice signal is detected by a time average calculation of the input voice signal of the input voice signal, and the silent section detecting means is Obtaining a period in which the value of the detected envelope is continuously lower than the set threshold value, when the period is longer than the preset shortest silent section, to detect the period as a silent section, A device for detecting and expanding a silent section of a characteristic voice. 4. The silent segment expansion means according to claim 1, wherein a value of an envelope curve of a voiced segment immediately before the silent segment,
Alternatively, a voice silent section detecting and expanding apparatus for expanding a silent section by setting a value proportional to a maximum value of an envelope. 5. The silent section of voice according to claim 4, wherein the silent section expanding unit repeatedly adds a central portion except for both ends of a start point and an end point of the silent section to the silent section. Detection and extension device. 6. The silent section expansion means according to claim 4, wherein the silent section expansion unit lowers the signal level of the silent section to a level lower than the signal level of the silent section in the input voice signal and adds the signal to the silent section. A device for detecting and expanding a silent section of a characteristic voice. 7. The envelope detecting means for detecting an envelope from an input voice signal converted into a digital signal according to claim 1, and a differentiating circuit for calculating a differential value of the envelope.
A delay circuit for delaying the input voice signal by a calculation time of the differentiating circuit, and the silent section detecting means sets a period during which the differential value of the envelope of the differentiating circuit changes from a negative peak value to a positive peak value. An apparatus for detecting and expanding a silent section of a voice, which is detected as a silent section. 8. The envelope detecting means according to claim 7, wherein the envelope of the input audio signal is detected by time-averaging the input audio signal of the input audio signal. Obtaining a period in which the differential value of the envelope by the differentiating circuit changes from a negative peak value to a positive peak value, and when the period is longer than a preset shortest silent section, this period is detected as a silent section. An apparatus for detecting and expanding a silent section of a voice, characterized by: 9. The differentiating circuit according to claim 7,
An apparatus for detecting and expanding a silent section of a voice, characterized in that a differential value is calculated by a convolution operation of the input voice signal converted into a digital signal and a window function having a point-symmetrical shape with respect to the origin. 10. The differentiating circuit according to claim 9, wherein a window function having a positive or negative slope is used,
An apparatus for detecting and expanding a silent section of a voice, characterized by calculating a differential value. 11. The apparatus for detecting and expanding a silent section of speech according to claim 9, wherein the differentiating circuit performs weighted addition using a non-linear window function. 12. The soundless segment expansion means according to claim 7, wherein the silent segment expansion unit expands the silent segment in proportion to the power of the negative peak value of the differential value immediately before the silent segment. Silent section detection and expansion device. 13. The silent section of voice according to claim 12, wherein the silent section expanding unit repeatedly adds a central portion except for both ends of a start point and an end point of the silent section to the silent section. Detection and extension device. 14. The silent section expansion means according to claim 12, wherein the level of the signal of the silent section is made lower than the level of the signal of the silent section in the input voice signal and added to the silent section. A device for detecting and expanding a silent section of a characteristic voice. 15. The envelope detecting means for detecting an envelope from an input audio signal converted into a digital signal, a first delay circuit for delaying information of the envelope, and the envelope according to claim 1. Information and delayed envelope information are input, each envelope information is weighted to a certain degree, and these are added to simulate the behavior of the OFF neuron, which corresponds to the rise and fall of the envelope. And a second delay circuit for delaying the input voice signal by the calculation time of the OFF neuron circuit, wherein the silent section detecting means outputs a positive value to the OFF neuron circuit. A silent interval detection / expansion device for voice, characterized in that a period from the output of the above to the output of the next negative value is detected as a silent interval. 16. The envelope detecting means according to claim 15, wherein the envelope of the input audio signal is detected by time-averaging the input audio signal of the input audio signal,
The silent section detecting means obtains a period from when the output from the OFF neuron circuit outputs a positive value to the next negative value, and the period is longer than a preset shortest silent section. In this case, the apparatus for detecting and expanding a silent section of voice, characterized in that the period is detected as a silent section. 17. The OFF neuron circuit according to claim 15, wherein the positive and negative are reversed, and a period from the output of a negative value to the output of the next positive value is detected as a silent section. A device for detecting and expanding a silent section of a voice. 18. The sound silence section according to claim 15, wherein the silence section expansion means expands the silence section in proportion to the power of the output of the OFF neuron circuit immediately before the silence section. Section detection and extension device. 19. The silent section of voice according to claim 18, wherein said silent section expanding means repeatedly adds a central portion except both ends of a start point and an end point of said silent section to said silent section. Detection and extension device. 20. The silent section expansion means according to claim 18, wherein the level of the signal in the silent section is made lower than the level of the signal in the silent section in the input voice signal and added to the silent section. A device for detecting and expanding a silent section of a characteristic voice. 21. The silent section detecting means according to claim 1, wherein the silent section is detected by using an input voice signal of one of two channels input to the ear of a person who listens to the input voice signal. A silent interval detecting / expanding device for voice, wherein the silent interval expanding means detects and expands a silent interval for both input audio signals of the two channels. 22. The silent section detecting means according to claim 21, wherein the one channel uses one of the input audio signals of the two channels, whichever has a larger input audio signal. A device for detecting and expanding silent sections of voice. 23. The soundless section detecting and expanding apparatus according to claim 21, wherein said silent section detecting means uses a channel on the side of the effective ear of a person who listens to said input audio signal as said one channel. apparatus. 24. The silent section detection means according to claim 1, wherein the silent section is detected by using an average value signal of two-channel input audio signals input to the ear of a person who listens to the input audio signal. A silent interval detecting / expanding device for audio, wherein the silent interval expanding means expands a silent interval for both input voice signals of the two channels. 25. A hearing aid is detected by detecting a silent section in an input audio signal and extending the signal of the detected silent section in proportion to the power of the input audio signal immediately before the silent section. A method for detecting and expanding a silent section of a voice, comprising: 26. The envelope according to claim 25, wherein an envelope is detected from the input voice signal converted into a digital signal, maximum and minimum values of the envelope are detected, and voice information of the input voice signal and the envelope are detected. At least the line information, the maximum value and the minimum value are stored, and a period in which the envelope becomes smaller than a threshold value set to any value between the maximum value and the minimum value is detected as a silent section, and the silent period is detected. A silent interval detecting and expanding method for voice, characterized in that the silent interval is expanded in proportion to the power of the input audio signal immediately before the interval. 27. The envelope according to claim 25, wherein an envelope is detected from an input voice signal converted into a digital signal, a differential value of the envelope is calculated, and the input voice signal is delayed by a calculation time of the differential value. Then, the period in which the differential value changes from the negative peak value to the positive peak value is detected as a silent section, and the silent section is expanded in proportion to the power of the negative peak value of the differential value immediately before the silent section. A method for detecting and expanding a silent section of a voice, comprising: 28. The envelope according to claim 25, wherein an envelope is detected from an input voice signal converted into a digital signal, information on the envelope is delayed, information on the envelope, and information on the delayed envelope. Is input, a certain weight is applied to the information of each envelope, these are added to simulate the operation of the OFF neuron, a signal corresponding to the rising and falling of the envelope is output, and the input voice signal is The period from the time when the OFF neuron outputs a positive value to the time when the OFF neuron outputs a next negative value is detected as a silent section by delaying by the calculation time of the OFF neuron, and the O immediately before the silent section is detected.
A silent interval detecting and expanding method for voice, characterized in that the silent interval is expanded in proportion to the output power of the FF neuron circuit. 29. A silent interval expansion means for detecting a silent interval in an input audio signal, and expanding the signal of the detected silent interval in proportion to the power of the input audio signal immediately before the silent interval, A telephone which assists hearing by using a voice silent period detection / expansion device including a silent period expansion amount changing means for changing the expansion amount of a silent period. 30. The method according to claim 29, further comprising: parameter storage means for storing a parameter set representing the expansion amount, and parameter selection means for selecting the previously used parameter set from the parameter storage means. And a telephone. 31. The telephone set according to claim 30, wherein the parameter selection means includes a frequency detection means for detecting a parameter set that is most frequently used among a plurality of parameter sets. 32. A silent interval expansion means for detecting a silent interval in an input audio signal, and expanding the signal of the detected silent interval in proportion to the power of the input audio signal immediately before the silent interval, A television receiver, characterized in that a hearing aid is assisted by using a device for detecting and expanding a silent section of a voice, comprising a silent section expansion amount changing means for changing the expansion amount of a silent section. 33. The method according to claim 32, further comprising parameter storage means for storing a parameter set representing the expansion amount and parameter selection means for selecting the parameter set used in the past from the parameter storage means. And a television receiver. 34. The television receiver according to claim 33, wherein the parameter selection means includes frequency detection means for detecting a parameter set having the highest frequency of use among a plurality of parameter sets. 35. A silent interval expansion means for detecting a silent interval in an input audio signal, and expanding the detected signal in the silent interval in proportion to the power of the input audio signal immediately before the silent interval, A radio receiver characterized in that it assists hearing by using a device for detecting and expanding a silent section of a voice, comprising: a unit for changing the expansion amount of a silent section. 36. The apparatus according to claim 35, further comprising parameter storage means for storing a parameter set representing the expansion amount, and parameter selection means for selecting the parameter set used in the past from the parameter storage means. And a radio receiver. 37. The radio receiver according to claim 36, wherein the parameter selecting means includes a frequency detecting means for detecting a parameter set having the highest frequency of use among a plurality of parameter sets.