JP5346230B2 - Speaking speed converter - Google Patents

Description

  The present invention relates to a speech speed conversion device that converts a speech speed of input speech and outputs the converted speech speed.

  Conventionally, various speech speed conversion devices that convert and output the speech speed of input speech have been proposed. Here, if the speech speed is simply slowed down, the output voice will be delayed, so the section containing sound (sounded section) and the section not containing sound (silent section) are discriminated. By delaying the input signal and compressing the input signal in the silent period, the output sound is prevented from being delayed (for example, see Patent Document 1).

JP-T-2006-77626

  However, in the conventional example that discriminates between the voiced section and the silent section as described above, since the silent section does not exist when noise is superimposed on the input signal, the input signal is always expanded, and the delay due to the speech speed conversion. There is a problem that increases. Therefore, by discriminating between a section (speech section) in which speech is included in the input signal and a section (non-speech section) in which speech is not included (delayed by non-speech), no delay occurs in the input signal on which noise is superimposed. It is conceivable to realize speech speed conversion. However, when the level of noise superimposed on the input signal is large, the speech section is misclassified as a non-speech section, and as a result, a part of the input speech is lost due to erroneous compression of the speech section. There was a fear.

  The present invention has been made in view of the above circumstances, and an object of the present invention is to provide a speech speed conversion device capable of appropriately converting speech speed even when noise is superimposed on input speech.

In order to achieve the above object, the first aspect of the present invention is a speech section discriminating section for discriminating between a speech section in which speech is included in an input signal and a non-speech section in which speech is not included, and the input signal is expanded. Alternatively, when the decompression / compression processing unit that compresses and outputs and the speech segment determination unit determines that the speech segment is a speech segment, the decompression / compression processing unit performs the decompression process and the speech segment determination unit determines that the segment is a non-speech segment A speech speed conversion processing determination unit that sometimes causes the compression / decompression processing unit to perform compression processing, and a noise level determination unit that determines a noise level included in the input signal, and the speech speed conversion processing determination unit is a noise level determination unit The decompression / compression processing unit determines whether or not the decompression process and the compression process can be performed according to the noise level determined by the sound level, and the noise level determination unit determines even if the speech segment determination unit determines that it is a non-speech segment. Be done Sound level is equal to or not to perform a compression process in the expansion compression processing unit when it is above a predetermined threshold.

According to the first aspect of the present invention, the speech speed conversion processing determination unit determines whether or not the expansion processing and the compression processing in the expansion / compression processing unit can be performed according to the noise level determined by the noise level determination unit. The speech speed can be appropriately converted even when is superimposed. Furthermore, even when the speech section determination unit determines that it is a non-speech section, the expansion / compression processing unit does not perform compression processing when the noise level determined by the noise level determination unit is greater than or equal to the threshold value. It is possible to prevent the output sound from being interrupted by the input sound being compressed by mistake.

In order to achieve the above object , a second aspect of the present invention provides a speech segment discriminating unit that discriminates between a speech segment that includes speech and a non-speech segment that does not include speech, and expands the input signal. Alternatively, when the decompression / compression processing unit that compresses and outputs and the speech segment determination unit determines that the speech segment is a speech segment, the decompression / compression processing unit performs the decompression process and the speech segment determination unit determines that the segment is a non-speech segment A speech speed conversion processing determination unit that sometimes causes the compression / decompression processing unit to perform compression processing, and a noise level determination unit that determines a noise level included in the input signal, and the speech speed conversion processing determination unit is a noise level determination unit The decompression processing in the decompression processing unit and whether or not to execute the compression process are determined according to the noise level determined by the sound level, and the noise level determination unit determines even if the speech segment determination unit determines that it is a speech segment. Noisy Level is equal to or not to perform the decompression processing to the decompression compression processing unit when it is above a predetermined threshold.

According to the second aspect of the present invention, the speech speed conversion processing determination unit determines whether or not the expansion processing and the compression processing in the expansion / compression processing unit can be performed according to the noise level determined by the noise level determination unit. The speech speed can be appropriately converted even when is superimposed. Furthermore, even if the speech segment determination unit determines that the speech segment is a speech segment, if the noise level determined by the noise level determination unit is greater than or equal to the threshold value, the expansion / compression processing unit does not perform the expansion process. It is possible to prevent the sound from being unnaturally stretched due to the influence of.

In order to achieve the above object , a third aspect of the present invention provides a voice section discriminating section that discriminates a voice section in which voice is included in an input signal and a non-voice section in which voice is not included, and expands the input signal. Alternatively, when the decompression / compression processing unit that compresses and outputs and the speech segment determination unit determines that the speech segment is a speech segment, the decompression / compression processing unit performs the decompression process and the speech segment determination unit determines that the segment is a non-speech segment A speech speed conversion processing determining unit that causes the decompression / compression processing unit to perform compression processing, and a noise level determining unit that determines a noise level included in the input signal. Even if the sound level is determined to be a non-speech segment, the second level of the noise level determined by the noise level determination unit is greater than or equal to a predetermined first threshold value and greater than the first threshold value. When it is less than the threshold, decompression and compression processing Without performing the compression process, further, extended compression when the voice segment discriminating unit noise level determined by the noise level determining unit even if is determined that the speech section is a second threshold or more It is characterized in that neither the decompression process nor the compression process is performed by the processing unit.

According to the third aspect of the present invention, the speech speed conversion process determination unit determines whether or not the expansion process and the compression process can be performed in the expansion / compression process unit according to the noise level determined by the noise level determination unit. The speech speed can be appropriately converted even when is superimposed. Furthermore, even when the speech segment determination unit determines that the speech segment is a non-speech segment, the noise level determined by the noise level determination unit is determined to be greater than or equal to the first threshold value and less than the second threshold value. When the decompression compression processing unit does not perform compression processing, the input speech is prevented from being interrupted by accidentally compressing the input speech, and the speech segment determination unit determines that it is a speech segment Even if the noise level determined by the noise level determination unit is determined to be greater than or equal to the second threshold value, the expansion / compression processing unit does not perform expansion processing, so that the sound is unnaturally expanded due to noise. Can be prevented.

According to a fourth aspect of the present invention, in any one of the first to third aspects, the noise level determination unit obtains the noise level using the absolute value of the amplitude of the input signal.

According to invention of Claim 4 , a noise level can be calculated | required comparatively easily.

According to a fifth aspect of the present invention, in any one of the first to third aspects, the noise level determination unit obtains a noise level using a frequency component of a non-voice band of the input signal.

According to the invention of claim 5 , the noise level can be obtained without depending on the level of the input voice.

A sixth aspect of the present invention includes a buffer unit that accumulates the input signal expanded or compressed by the expansion / compression processing unit according to any one of the first to fifth aspects of the present invention. Even when the section discriminating unit discriminates the voice section, the decompression / compression processing unit is not allowed to perform the decompression process when the free capacity of the buffer unit is equal to or less than a predetermined lower limit value.

According to the sixth aspect of the present invention, it is possible to prevent malfunction caused by overflow of the buffer section and overwriting of output sound in another memory area.

The invention of claim 7 is the invention of any one of claims 1 to 6 , further comprising a buffer unit that accumulates the input signal expanded or compressed by the expansion / compression processing unit, and the speech speed conversion processing determination unit Even when the section determination unit determines that the section is a non-speech section, the decompression / compression processing unit is not allowed to perform compression processing when the free space in the buffer unit is equal to or greater than a predetermined upper limit value.

According to the seventh aspect of the present invention, it is possible to prevent malfunction caused by the buffer section underflowing (underrun) and overwriting the output sound in another memory area.

The invention according to claim 8 is the invention according to any one of claims 1 to 7 , further comprising a buffer unit that accumulates an input signal expanded or compressed by the expansion / compression processing unit, and the speech speed conversion process determining unit is expanded The expansion rate when the compression processing unit performs the expansion process is increased or decreased according to the free capacity of the buffer unit.

According to the eighth aspect of the present invention, it is possible to prevent a voice that has not been subjected to speech speed conversion from being suddenly output when there is no more free space in the buffer unit.

  According to the present invention, speech speed can be appropriately converted even when noise is superimposed on input speech.

It is a block diagram which shows embodiment of this invention. (A) is a waveform diagram of the absolute value of the input speech on which noise is not superimposed, and (b) is a waveform diagram of the absolute value of the input speech on which noise is superimposed. It is operation | movement explanatory drawing same as the above.

  Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

  As shown in FIG. 1, the speech speed conversion apparatus according to the present embodiment includes a speech section determination unit 1 that determines a speech section in which speech is included in an input signal and a non-speech section in which speech is not included, and an input signal. The decompression / compression processing unit 2 that decompresses or compresses and outputs the decompression / compression processing unit 2 causes the decompression / compression processing unit 2 to perform decompression processing when the speech segment determination unit 1 determines that it is a speech segment, and the speech segment determination unit 1 performs non-speech A speech speed conversion processing determination unit 3 that causes the expansion / compression processing unit 2 to perform compression processing when it is determined to be a section, a noise level determination unit 4 that determines a noise level included in an input signal, and an expansion / compression processing unit 2 And a buffer unit 5 for accumulating the input signal expanded or compressed in (1). Each of these units is realized by causing a DSP (Digital Signal Processor) to execute a predetermined program, and can be obtained by, for example, A / D converting an analog acoustic signal collected by a microphone. A digital signal becomes an input signal.

  The voice section discriminating unit 1 is a conventionally known one, discriminates the voice section of the input signal and outputs the discrimination result to the speech speed conversion processing judgment unit 3.

  The decompression / compression processing unit 2 performs time-axis companding (compression and decompression) processing on the input signal and adjusts the speech speed of the speech represented by the speech signal. An algorithm called PICOLA that performs speech speed conversion is performed by calculating the correlation, regarding the frame length with the highest correlation as the period of the speech, and inserting or deleting the waveform in that period.

  The noise level determination unit 4 is composed of a digital filter having a characteristic that the rise is gradual and the fall is steep, that is, a response characteristic having a relatively large rise time constant and a relatively small fall time constant. The absolute value of the amplitude of the signal is used to determine (estimate) the level of noise (noise level) that is constantly present in the input signal. Comparing the case where the input signal is speech only and no noise is superimposed as shown in FIG. 2A, and the case where the input signal is speech and noise is superimposed as shown in FIG. While the amplitude waveform of speech changes relatively abruptly, the amplitude waveform of noise changes relatively slowly, so the response with a relatively large rise time constant and a relatively small fall time constant The noise level can be determined by removing the amplitude waveform of the voice from the amplitude waveform of the input signal by a digital filter having characteristics. FIG. 2 shows the waveform of the absolute value of the input signal. However, the noise level determination unit 4 can perform discrete fast Fourier transform to determine the noise level using the frequency component of the non-speech frequency band of the input signal.

  The speech speed conversion processing determination unit 3 determines whether or not the expansion processing and the compression processing in the expansion / compression processing unit 2 are executable according to the noise level determined by the noise level determination unit 4. Specifically, in the speech speed conversion process determination unit 3, the noise level determined by the noise level determination unit 4 is greater than or equal to the threshold value even when the voice segment determination unit 1 determines that it is a non-speech segment. In some cases, the decompression / compression processing unit 2 is not allowed to perform compression processing. That is, when the noise level is equal to or higher than the threshold value, there is a high possibility that the voice segment determination unit 1 erroneously discriminates the voice segment as a non-speech segment. In such a case, the decompression compression processing unit 2 performs compression processing. By not performing it, it is possible to appropriately convert the speech speed even when noise is superimposed on the input voice. Specifically, it is possible to prevent the output voice from being interrupted due to erroneous compression of the input voice.

  Further, when the speech level conversion processing determination unit 3 determines that the noise level determined by the noise level determination unit 4 is equal to or greater than a predetermined threshold even when the speech segment determination unit 1 determines that the speech segment is a speech segment. If the expansion / compression processing unit 2 is not allowed to perform the expansion process, it is possible to prevent the sound from being unnaturally expanded due to the influence of noise.

Alternatively, even if the speech speed conversion processing determination unit 3 is determined as a non-speech segment by the speech segment determination unit 1, the noise level determined by the noise level determination unit 4 is a predetermined first threshold value. If it is above and less than the second threshold value that is greater than the first threshold value, the decompression / compression processing unit 2 does not perform compression processing, and the speech segment determination unit 1 determines that it is a speech segment. Even if the noise level determined by the noise level determination unit 4 is equal to or higher than the second threshold value, if the expansion / compression processing unit 2 is not allowed to perform either expansion processing or compression processing, It is possible to prevent the output sound from being interrupted by the input sound being compressed by mistake, and to prevent the sound from being unnaturally expanded due to the influence of noise.

  By the way, the input signal after being decompressed or compressed by the decompression / compression processing unit 2 is temporarily stored in the buffer unit 5 and then output as an output signal subjected to speech speed conversion. Then, the buffer unit 5 overflows, and there is a risk of malfunction due to the output signal (audio data) being overwritten in another memory area. Similarly, if the voice section of the input signal becomes shorter, the buffer unit 5 may underflow (underrun), and a malfunction may occur due to overwriting of the output signal (voice data) in another memory area.

  Therefore, in the speech speed conversion process determination unit 3 according to the present embodiment, even when the voice segment determination unit 1 determines that the voice segment is a voice segment, if the free capacity of the buffer unit 5 is equal to or less than a predetermined lower limit value, the expansion / compression process is performed. By not allowing the unit 2 to perform the decompression process, the buffer unit 5 is prevented from overflowing, and the free space of the buffer unit 5 has a predetermined upper limit even when the voice segment determination unit 1 determines that it is a non-speech segment. When the value is greater than or equal to the value, the decompression / compression processing unit 2 is not allowed to perform compression processing, thereby preventing underflow (underrun) of the buffer unit 5.

  In addition, when the expansion rate in the expansion / compression processing unit 2 is fixed, the free capacity of the buffer unit 5 decreases as the audio section becomes longer, and the voice whose speech speed is not suddenly converted when the free capacity of the buffer unit is exhausted. May be output. On the other hand, in the speech speed conversion process determination unit 3 of the present embodiment, the expansion rate when the expansion / compression processing unit 2 performs the expansion process is determined according to the free capacity of the buffer unit 5 (for example, in proportion). By increasing / decreasing, it is possible to prevent the voice that has not been subjected to the speech speed conversion from being suddenly output due to the empty capacity of the buffer unit 5 being lost. For example, as shown in FIG. 3, when the input voice of the sentence “Sho Nerugi is as much as possible” is converted (expanded), the free space of the buffer unit 5 remains at the beginning of the sentence. Therefore, the expansion rate may be increased, and the expansion rate may be decreased as the free space of the buffer unit 5 decreases near the end of the sentence.

  Here, the speech speed conversion apparatus according to the present embodiment is suitable for a so-called hands-free type intercom device. In other words, in an intercom system that performs a hands-free call (speech call) between an intercom device installed outside the dwelling unit (entrance handset or lobby intercom) and an intercom device installed inside the dwelling unit (master unit) In many cases, outdoor noise is superimposed on the sound collected by the microphone of the machine or lobby intercom. Therefore, by installing the speech speed conversion device of the present embodiment in the master unit, it is possible to appropriately convert the speech speed even when noise is superimposed on the input speech (speech input from the front door unit or lobby intercom). A remarkable effect is obtained.

DESCRIPTION OF SYMBOLS 1 Speech section determination part 2 Decompression compression processing part 3 Speech rate conversion process determination part 4 Noise level determination part 5 Buffer part

Claims (8)

A speech segment determination unit that determines a speech segment in which speech is included in an input signal and a non-speech segment in which speech is not included, a decompression / compression processing unit that decompresses or compresses the input signal, and a speech segment determination Speed conversion that causes the decompression / compression processing unit to perform decompression processing when the speech section is determined to be a speech segment and causes the decompression / compression processing unit to perform compression processing when the speech segment determination unit determines that it is a non-speech segment A processing determination unit, and a noise level determination unit that determines a noise level included in the input signal,
The speech speed conversion processing determination unit determines whether or not the expansion processing and the compression processing in the expansion / compression processing unit can be performed according to the noise level determined by the noise level determination unit, and the speech section determination unit determines that the non-speech section is determined. A speech speed converting apparatus characterized in that the compression / decompression processing unit does not perform compression processing when the noise level determined by the noise level determination unit is equal to or greater than a predetermined threshold even when the noise level is determined . A speech segment determination unit that determines a speech segment in which speech is included in an input signal and a non-speech segment in which speech is not included, a decompression / compression processing unit that decompresses or compresses the input signal, and a speech segment determination Speed conversion that causes the decompression / compression processing unit to perform decompression processing when the speech section is determined to be a speech segment and causes the decompression / compression processing unit to perform compression processing when the speech segment determination unit determines that it is a non-speech segment A processing determination unit, and a noise level determination unit that determines a noise level included in the input signal,
Speech speed conversion processing determination unit is configured to determine whether to execute expansion processing and compression processing in the decompression compression processor according to the noise level determining noise level determining unit, speech segment determination section to determine the voice interval A speech speed converting apparatus characterized in that even when the noise level determined by the noise level determination unit is equal to or greater than a predetermined threshold value, the expansion / compression processing unit does not perform the expansion processing. A speech segment determination unit that determines a speech segment in which speech is included in an input signal and a non-speech segment in which speech is not included, a decompression / compression processing unit that decompresses or compresses the input signal, and a speech segment determination Speed conversion that causes the decompression / compression processing unit to perform decompression processing when the speech section is determined to be a speech segment and causes the decompression / compression processing unit to perform compression processing when the speech segment determination unit determines that it is a non-speech segment A processing determination unit, and a noise level determination unit that determines a noise level included in the input signal,
Speech speed conversion processing determination unit, and Ri first der than the threshold noise level is given even when it is determined by the noise level determining unit which speech segment determination unit is to determine the non-speech section Even if it is less than the second threshold value, which is larger than the first threshold value, the decompression / compression processing unit does not perform the compression process, and further, even if the speech segment determination unit determines that it is a speech segment An apparatus for converting speech speed according to claim 1, wherein when the noise level determined by the noise level determination unit is equal to or higher than a second threshold value, the expansion / compression processing unit does not perform either expansion processing or compression processing . The speech speed conversion apparatus according to any one of claims 1 to 3, wherein the noise level determination unit obtains a noise level using an absolute value of an amplitude of an input signal . The speech speed conversion apparatus according to any one of claims 1 to 3 , wherein the noise level determination unit obtains a noise level using a frequency component of a non-voice band of the input signal. A buffer unit for storing the input signal expanded or compressed by the expansion / compression processing unit;
The speech rate conversion processing determination unit does not cause the expansion / compression processing unit to perform the expansion process when the free space of the buffer unit is equal to or less than a predetermined lower limit value even when the speech section determination unit determines that the speech section is a speech section. The speech rate conversion apparatus according to any one of claims 1 to 5 , wherein A buffer unit for storing the input signal expanded or compressed by the expansion / compression processing unit;
Speech speed conversion processing determination unit, to perform a compression process in the expansion compression processing unit when the voice segment discriminating unit free space of the buffer unit even if is determined that the non-voice section is equal to or greater than a predetermined upper limit value The speech rate conversion apparatus according to claim 1, wherein the speech rate conversion apparatus is not provided. A buffer unit for storing the input signal expanded or compressed by the expansion / compression processing unit;
The speech rate conversion processing determination unit increases or decreases the expansion rate when the expansion / compression processing unit performs expansion processing according to the free capacity of the buffer unit. The speech rate conversion device described .

Images