JP4736699B2 - Audio signal compression apparatus, audio signal restoration apparatus, audio signal compression method, audio signal restoration method, and program - Google Patents

Description

  The present invention relates to an audio signal compression device, an audio signal restoration device, an audio signal compression method, an audio signal restoration method, and a program.

In recent years, a speech synthesis method for converting text data into speech has been used in the field of car navigation and the like.
In speech synthesis, for example, a word included in a sentence represented by text data, a phrase, and a dependency relationship between phrases are specified, and how to read the sentence is specified based on the specified word, phrase, and dependency relationship. The phoneme waveform, duration, and pitch (fundamental frequency) patterns that make up the speech are determined based on the phonetic character string that represents the specified reading. Based on the determination result, The waveform is determined, and a sound having the determined waveform is output.

  In the speech synthesis method described above, in order to specify a speech waveform, a speech dictionary in which speech data representing the speech waveform is accumulated is searched. In order for the synthesized speech to be natural, the speech dictionary must accumulate an enormous number of speech data.

  In addition, when this method is applied to a device that is required to be downsized, such as a car navigation device, generally, a storage device that stores a speech dictionary used by the device needs to be downsized. If the size of the storage device is reduced, it is generally inevitable to reduce the storage capacity.

Therefore, in order to store a phoneme dictionary including a sufficient amount of audio data even in a storage device with a small storage capacity, it is possible to compress the audio data and reduce the data capacity per audio data. (For example, refer to Patent Document 1).
Special Table 2000-502539

  However, using entropy coding techniques (specifically, arithmetic coding, Huffman coding, etc.) that compress data by paying attention to the regularity of the data, audio data representing the voice uttered by a person is converted. In the case of compression, although sound data representing a voice uttered by a person has a certain degree of regularity and can be efficiently compressed, a component that does not originate from a voice uttered by a person (for example, a component that represents a sound emitted by a musical instrument) Etc.) is not necessarily clearly periodic as a whole, and compression efficiency is low.

  Further, when entropy coding is performed on voice data representing voice uttered by a person, pitch fluctuation has also been a problem. The pitch is easily affected by human emotions and consciousness and is a period that can be regarded as being constant to some extent, but in reality it causes subtle fluctuations. Therefore, when the same speaker utters the same word (phoneme) for a plurality of pitches, the pitch interval is usually not constant. Therefore, there are many cases where accurate regularity is not observed even in a waveform representing one phoneme, and for this reason, compression efficiency by entropy coding is often lowered.

  The present invention has been made in view of the above circumstances, and an audio signal compression device, an audio signal compression method, and an audio signal compression method for efficiently compressing the data capacity of data including a component representing a voice uttered by a person, and It is an object of the present invention to provide a program, and to provide an audio signal restoration device, an audio signal restoration method, and a program for restoring data compressed by such an audio signal compression device and an audio signal compression method.

In order to achieve the above object, an audio signal compression apparatus according to the first aspect of the present invention provides:
Subband extraction means for generating a subband signal representing a temporal change in the intensity of the fundamental frequency component and the harmonic component of the audio signal to be compressed representing the waveform of the audio;
When it is determined that the amplitude of the pitch signal obtained by filtering the audio signal does not reach a predetermined amount, the subband signal has a continuous component having a periodicity that meets a predetermined criterion, and A random component corresponding to the subband signal excluding the continuous component is separated, and the subband signal is treated as the continuous component when it is determined that the amplitude of the pitch signal has reached the predetermined amount. Component separation means;
Encoding means for performing entropy encoding or linear predictive encoding on the continuous components ;
Characterized the Ruco equipped with.

An audio signal compression apparatus according to the second aspect of the present invention provides:
By acquiring the audio signal to be compressed representing the waveform of the audio, and dividing the audio signal into a plurality of intervals corresponding to the unit pitch of the audio, the phases of these intervals are made substantially the same, Audio signal processing means for processing the signal into a pitch waveform signal;
Subband extraction means for generating a subband signal representing a temporal change in intensity of the fundamental frequency component and the harmonic component of the pitch waveform signal;
When it is determined that the amplitude of the pitch signal obtained by filtering the audio signal does not reach a predetermined amount, the subband signal has a continuous component having a periodicity that meets a predetermined criterion, and A random component corresponding to the subband signal excluding the continuous component is separated, and the subband signal is treated as the continuous component when it is determined that the amplitude of the pitch signal has reached the predetermined amount. Component separation means;
Encoding means for performing entropy encoding or linear predictive encoding on the continuous components ;
Characterized the Ruco equipped with.

  The encoding means may perform entropy encoding on the result of nonlinear quantization of the continuous component and / or the result of nonlinear quantization of the random component.

  The encoding unit may generate data indicating a quantization characteristic of the nonlinear quantization.

  The encoding means determines a quantization characteristic of the nonlinear quantization based on a data amount of a continuous component and / or a random component that have been entropy-encoded in the past, and matches the determined quantization characteristic. Nonlinear quantization may be performed.

An audio signal restoration device according to the third aspect of the present invention is
An entropy code is extracted by extracting continuous components having a periodicity that meets a predetermined standard from subband signals representing temporal changes in the fundamental frequency components and harmonic component intensities of audio signals to be compressed that represent speech waveforms. Decoding means for obtaining an input signal corresponding to a signal that has been subjected to normalization or linear predictive coding, and restoring the continuous component by decoding the input signal;
If the random component corresponding to the restored continuous component minus from said sub-band signal is present, by obtains the random component, adds the random component to the connected component, the sub-band A subband signal restoration unit that restores a signal and treats the restored continuous component as the subband signal when the random component does not exist ;
And voice signal restoring means for restoring the audio signal of the compressed object based on the reconstructed subband signals,
Characterized the Ruco equipped with.

An audio signal compression method according to the fourth aspect of the present invention includes:
An audio signal compression method executed by an audio signal compression apparatus having a processor,
The processor generates a subband signal representing a temporal change in intensity of a fundamental frequency component and a harmonic component of an audio signal to be compressed representing an audio waveform;
When the processor determines that the amplitude of the pitch signal obtained by filtering the audio signal does not reach a predetermined amount, the processor has a periodicity that meets a predetermined criterion from the subband signal. When the component and a random component corresponding to the sub-band signal obtained by removing the continuous component are separated and it is determined that the amplitude of the pitch signal has reached the predetermined amount, the sub-band signal is Treated as a continuous component,
The processor performs entropy coding or linear predictive coding on the continuous components;
It is characterized by that.

An audio signal compression method according to the fifth aspect of the present invention is
An audio signal compression method executed by an audio signal compression apparatus having a processor,
When the processor acquires an audio signal to be compressed representing a waveform of an audio and divides the audio signal into a plurality of intervals corresponding to a unit pitch of the audio, the phases of these intervals are made substantially the same. By processing the audio signal into a pitch waveform signal,
The processor generates a subband signal representing a temporal change in intensity of a fundamental frequency component and a harmonic component of the pitch waveform signal;
When the processor determines that the amplitude of the pitch signal obtained by filtering the audio signal does not reach a predetermined amount, the processor has a periodicity that meets a predetermined criterion from the subband signal. When the component and the random component corresponding to the sub-band signal obtained by removing the continuous component are separated and it is determined that the amplitude of the pitch signal has reached a predetermined amount, the sub-band signal is Treat as an ingredient,
The processor performs entropy coding or linear predictive coding on the continuous components;
It is characterized by that.

An audio signal restoration method according to the sixth aspect of the present invention includes:
An audio signal restoration method executed by an audio signal restoration device having a processor,
The processor extracts a continuous component having a periodicity enough to meet a predetermined standard from subband signals representing temporal changes in intensity of fundamental frequency components and harmonic components of a compression target speech signal representing a speech waveform. To obtain an input signal corresponding to the one subjected to entropy coding or linear prediction coding, and to restore the continuous component by decoding the input signal,
Wherein the processor, when the random component corresponding to the restored continuous component minus from said sub-band signal is present, acquires the random component, by adding the random component to the continuous component The subband signal is restored, and when the random component is not present, the restored continuous component is treated as the subband signal,
The processor restores the audio signal to be compressed based on the restored subband signal;
It is characterized by that.

A program according to the seventh aspect of the present invention is
Computer
Subband extraction means for generating a subband signal representing a temporal change in intensity of a fundamental frequency component and a harmonic component of an audio signal to be compressed representing an audio waveform ;
When it is determined that the amplitude of the pitch signal obtained by filtering the audio signal does not reach a predetermined amount, the subband signal has a continuous component having a periodicity that meets a predetermined criterion, and A random component corresponding to the subband signal excluding the continuous component is separated, and the subband signal is treated as the continuous component when it is determined that the amplitude of the pitch signal has reached the predetermined amount. Component separation means ,
Encoding means for performing entropy encoding or linear predictive encoding on the continuous components ;
Characterized in that it is intended to function as a.

A program according to the eighth aspect of the present invention is
Computer
By acquiring the audio signal to be compressed representing the waveform of the audio, and dividing the audio signal into a plurality of intervals corresponding to the unit pitch of the audio, the phases of these intervals are made substantially the same, Audio signal processing means for processing a signal into a pitch waveform signal ;
Subband extraction means for generating a subband signal representing a temporal change in intensity of the fundamental frequency component and the harmonic component of the pitch waveform signal ;
When it is determined that the amplitude of the pitch signal obtained by filtering the audio signal does not reach a predetermined amount, the subband signal has a continuous component having a periodicity that meets a predetermined criterion, and A random component corresponding to the subband signal excluding the continuous component is separated, and the subband signal is treated as the continuous component when it is determined that the amplitude of the pitch signal has reached the predetermined amount. Component separation means ,
Encoding means for performing entropy encoding or linear predictive encoding on the continuous components ;
Characterized in that it is intended to function as a.

A program according to the ninth aspect of the present invention is
Computer
An entropy code is extracted by extracting continuous components having a periodicity that meets a predetermined standard from subband signals representing temporal changes in the fundamental frequency components and harmonic component intensities of audio signals to be compressed that represent speech waveforms. A decoding unit that obtains an input signal corresponding to a signal that has been subjected to normalization or linear predictive coding, and restores the continuous component by decoding the input signal ;
If the random component corresponding to the restored continuous component minus from said sub-band signal is present, by obtains the random component, adds the random component to the connected component, the sub-band A subband signal restoration unit that restores a signal and treats the restored continuous component as the subband signal when the random component is not present ;
Voice signal restoring means for restoring the audio signal of the compressed object based on the reconstructed subband signals,
Characterized in that it is intended to function as a.

  According to the present invention, an audio signal compression device, an audio signal compression method, and a program for efficiently compressing a data capacity of data including a component representing audio uttered by a person are realized. An audio data restoration device, an audio data restoration method, and a program for restoring the data compressed by the audio signal compression device and the audio signal compression method are realized.

Embodiments of the present invention will be described below with reference to the drawings.
(First embodiment)
FIG. 1 is a diagram showing the configuration of an audio data compression system according to the first embodiment of the present invention. As shown in the figure, this audio data compression system includes a recording medium drive device (a flexible disk drive or a CD-R) that reads data recorded on a recording medium (for example, a flexible disk or a CD-R (Compact Disc-Recordable)). (ROM drive etc.) SMD and computer C1 connected to the recording medium drive device SMD.

  As shown in the figure, a computer C1 includes a processor composed of a CPU (Central Processing Unit) and a DSP (Digital Signal Processor), a volatile memory composed of a RAM (Random Access Memory), etc., and a nonvolatile memory composed of a hard disk device and the like. And an input unit including a keyboard, a display unit including a liquid crystal display, and a serial communication control unit configured to control serial communication with the outside, including a USB (Universal Serial Bus) interface circuit.

  The computer C1 stores an audio data compression program in advance, and performs the processing described later by executing the audio data compression program.

(First Embodiment: Operation)
Next, the operation of this audio data compression system will be described with reference to FIGS. 2 and 3 are diagrams showing an operation flow of the audio data compression system of FIG.

  When the user sets a recording medium on which audio data representing an audio waveform is recorded in the recording medium drive device SMD and instructs the computer C1 to start the audio data compression program, the computer C1 performs processing of the audio data compression program. To start.

  Then, first, the computer C1 reads audio data from the recording medium via the recording medium drive device SMD (FIG. 2, step S101). Note that the audio data has, for example, a PCM (Pulse Code Modulation) modulated digital signal format, and represents audio sampled at a constant cycle sufficiently shorter than the audio pitch.

  Next, the computer C1 generates filtered voice data (pitch signal) by filtering the voice data read from the recording medium (step S102). The pitch signal is assumed to be digital data having a sampling interval substantially the same as the sampling interval of audio data.

  The computer C1 determines the characteristics of the filtering performed to generate the pitch signal by performing feedback processing based on the pitch length described later and the time when the instantaneous value of the pitch signal becomes 0 (time when zero crossing). To do.

  That is, the computer C1 identifies the fundamental frequency of the voice represented by the voice data by performing, for example, cepstrum analysis or analysis based on the autocorrelation function on the read voice data, and the absolute value of the reciprocal of the fundamental frequency. (That is, the pitch length) is obtained (step S103). (Alternatively, the computer C1 specifies two fundamental frequencies by performing both cepstrum analysis and analysis based on the autocorrelation function, and obtains the average of the absolute values of the reciprocals of these two fundamental frequencies as the pitch length. May be.)

  For cepstrum analysis, specifically, the intensity of the read audio data is first converted to a value substantially equal to the logarithm of the original value (the base of the logarithm is arbitrary), and the value is converted. The spectrum (ie, cepstrum) of the audio data is obtained by a fast Fourier transform method (or any other method that generates data representing the result of Fourier transform of discrete variables). Then, the minimum value of the frequencies giving the maximum value of the cepstrum is specified as the fundamental frequency.

  On the other hand, as the analysis based on the autocorrelation function, specifically, the autocorrelation function r (l) represented by the right side of Formula 1 is first specified using the read audio data. Then, a minimum value exceeding a predetermined lower limit value is specified as a fundamental frequency among frequencies giving a maximum value of a function (periodogram) obtained as a result of Fourier transform of the autocorrelation function r (l).

  On the other hand, the computer C1 specifies the timing at which the time when the pitch signal crosses zero (step S104). Then, the computer C1 determines whether or not the zero-cross cycle and the pitch length of the pitch signal are different from each other by a predetermined amount or more (step S105), and if it is determined that they are not different, the reciprocal of the zero-cross cycle is the center. It is assumed that the above-described filtering is performed with the characteristics of the band-pass filter that has the frequency (step S106). On the other hand, if it is determined that they differ by a predetermined amount or more, the above-described filtering is performed with the characteristics of a bandpass filter having the reciprocal of the pitch length as the center frequency (step S107). In any case, it is desirable that the filtering pass band width is such that the upper limit of the pass band always falls within twice the fundamental frequency of the voice represented by the voice data.

  Next, the computer C1 divides the audio data read from the recording medium at the timing when the boundary of the unit period (for example, one period) of the generated pitch signal comes (specifically, the timing at which the pitch signal crosses zero) (step S1). S108). Then, for each of the sections that can be divided, the correlation between the variously changed phases of the audio data in this section and the pitch signal in this section is obtained, and the phase of the audio data when the correlation becomes the highest is obtained. The phase of the audio data in this section is specified (step S109). Then, the respective sections of the audio data are phase-shifted so that they have substantially the same phase (step S110).

  Specifically, the computer C1 changes, for each section, for example, the value cor represented by the right side of Formula 2 and the value of φ representing the phase (where φ is an integer of 0 or more). Ask for each case. Then, the value ψ of φ that maximizes the value cor is specified as a value representing the phase of the audio data in this section. As a result, the value of the phase having the highest correlation with the pitch signal is determined for this section. Then, the computer C1 shifts the audio data in this section by (−Ψ).

  Data obtained by phase-shifting audio data as described above is pitch waveform data. An example of the waveform represented by the pitch waveform data is shown in FIG. Among the waveforms of the audio data before phase shift shown in FIG. 4A, two sections indicated as “# 1” and “# 2” are affected by pitch fluctuations as shown in FIG. 4B. Have different phases. On the other hand, in the sections # 1 and # 2 of the waveform represented by the phase-shifted audio data (that is, the pitch waveform data), as shown in FIG. It's all there. Further, as shown in FIG. 4A, the value of the start point of each section is a value close to zero.

  Note that the time length of the section is preferably about one pitch. As the section becomes longer, the number of samples in the section increases and the amount of pitch waveform data increases, or the sampling interval increases and the voice represented by the pitch waveform data becomes inaccurate.

  Next, the computer C1 interpolates pitch waveform data (step S111). That is, interpolation data representing values for interpolating between samples of pitch waveform data is generated and added to the pitch waveform data, thereby generating pitch waveform data after interpolation.

Next, the computer C1 resamples (resamples) each section of the pitch waveform data after interpolation. In addition, sample number data that is data indicating the original number of samples in each section is also generated (step S112). Note that the computer C1 performs resampling so that the number of samples in each section of the pitch waveform data is substantially equal to each other, and is equally spaced within the same section.
Assuming that the sampling interval of the audio data read from the recording medium is known, the sample number data functions as information representing the original time length of the unit pitch of the audio data.

  Next, the computer C1 generates a subband data group by performing orthogonal transformation such as DCT (Discrete Cosine Transform) on the resampled pitch waveform data (step S113). The subband data group includes data (0th subband data) representing the temporal change in intensity of the fundamental frequency component of the voice represented by the resampled pitch waveform data, and n harmonics (n is a natural number) of this voice. It is composed of n pieces of data (1st to nth subband data) representing the time change of the intensity of the wave component. (Therefore, the subband data represents the intensity of the fundamental frequency component (or harmonic component) in the form of a direct current signal when there is no temporal change in the intensity of the fundamental frequency component (or harmonic component) of the voice.)

Next, the computer C1 determines whether or not the amplitude of the pitch signal generated in step S102 has reached a predetermined amount (step S114). If it is determined that the amplitude has not reached, the subband generated in step S113 is determined. By filtering each of the subband data included in the data group, data representing a component having a periodicity stronger than a certain degree among each subband data (hereinafter referred to as continuous component data) is generated, Data representing a component obtained by removing the continuous component from the subband data (hereinafter referred to as random component data) is also generated (step S115), and the process proceeds to step S117.
(Hereinafter, the continuous component data separated from the k-th (k is an integer from 0 to n)) sub-band data is referred to as k-th continuous component data, and is separated from the k-th sub-band data. The random component data is called the kth random component data.)

  On the other hand, if it is determined in step S114 that the amplitude of the pitch signal has reached the predetermined amount, the computer C1 determines to treat the kth subband data as it is as the kth continuous component data (step S116). The process moves to step S117.

In general, speech uttered by humans contains a lot of periodic pitch components, while other sounds (such as sounds generated by musical instruments) contain a lot of periodic components. I can't. Therefore, it can be considered that the above-mentioned continuous component data represents a component caused by a voice uttered by a person in the subband data, while the above-mentioned random component data contains a component not caused by a voice uttered by a person. It can be seen as a representation.
The process performed by the computer C1 in step S114 is as follows: “Non-existence of components that do not originate in the speech uttered by a person in the subband data group and ignore all the components in the subband data group as components of the speech uttered by the person. It is equivalent to the process of determining whether or not it can be treated as an expression of the signal, and a component that is not caused by the voice uttered by a person cannot be ignored (specifically, the amplitude of the pitch signal reaches a predetermined amount). In other words, the subband data corresponds to a process of separating the subband data into a component that is considered to be caused by speech uttered by a person and a component that is considered not to be caused.

  Next, the computer C1 uses the (n + 1) continuous component data obtained in step S115 or S116 and the (n + 1) random component data obtained in step S115 to (n + 1) nonlinear quantization continuous data. Component data and (n + 1) nonlinear quantized random component data are generated (step S117), and data including (n + 1) nonlinear quantized continuous component data obtained in step S117 and (n + 1) nonlinear quanta Each of the data including the randomized component data is entropy-encoded to generate continuous component compressed data and random component compressed data, which will be described later (step S118), and the compression rate of the continuous component data (that is, (n + 1) continuous data). For the total amount of component data and (n + 1) random component data, The ratio of the total amount of component compressed data and random component compressed data) and a predetermined target value are determined (step S119). Repeat until the target value is reached.

  Specifically, the computer C1, for example, first in step S117, a value obtained by applying nonlinear compression to the instantaneous value of the waveform represented by each of the (n + 1) continuous component data obtained in step S115 or S116 ( Specifically, for example, a total of (n + 1) data corresponding to the quantized value obtained by substituting an instantaneous value into an upward convex function is converted into the above-described (n + 1) nonlinear quantization. Generated as continuous component data. Further, in step S117, the computer C1 corresponds to the quantized value obtained by applying the nonlinear compression to the instantaneous value of the waveform represented by each of the (n + 1) random component data obtained in step S115. A total of (n + 1) pieces of data are generated as the above-described (n + 1) pieces of nonlinear quantized random component data.

The compression characteristic of the nonlinear compression performed by the computer C1 in step S117 (that is, the correspondence relationship between the pre-compression value and the post-compression value of the instantaneous value) is based on the result of the process of step S119 executed most recently. The computer C1 determines. Specifically, if it is determined that the compression rate obtained in step S119 is larger than the target value, the compression characteristics are determined so that the compression rate is smaller than the current value. On the other hand, if it is determined that the obtained compression rate is smaller than the target value, the compression characteristics are determined so that the compression rate is greater than the current value. However, if the process of step S119 has not been executed yet, compression is performed using a predetermined initial characteristic as a compression characteristic.
In step S117, the computer C1 creates compression characteristic data indicating the determined compression characteristic.

  A more specific example of the procedure for determining the compression characteristics in step S117 will be described. The computer C1, for example, the function global_gain (xi) included in the right side of Equation 3 is the result of the process of step S119 executed most recently. Determine based on. Then, the instantaneous value of each continuous component data and each random component data after nonlinear compression is changed to a value that is substantially equal to a value obtained by quantizing the function Xri (xi) shown on the right side of Equation 3. To perform nonlinear quantization. On the other hand, the computer C1 creates data representing the determined function global_gain (xi) as the above-described compression characteristic data.

(Expression 3) Xri (xi) = sgn (xi) · | xi | ^4/3 · 2 ^{{global_gain (xi)} / 4}
(Where sgn (α) = (α / | α |), xi is the instantaneous value of the waveform of the continuous component data, and global_gain (xi) is a function of xi for setting the full scale)

  Next, in step S118, the computer C1 entropy-encodes (n + 1) pieces of nonlinear quantized continuous component data and compression characteristic data obtained in step S117 (specifically, for example, an arithmetic code) Alternatively, continuous component compressed data is generated by converting the data into a Huffman code. In step S118, the computer C1 entropy-codes the (n + 1) nonlinear quantized random component data obtained in step S117, thereby generating random component compressed data.

  Next, in step S119, the computer C1 determines the continuous component compressed data obtained in step S118 for the total amount of (n + 1) continuous component data and (n + 1) random component data obtained in step S114. And the ratio of the total amount of random component compressed data is obtained as a compression rate, and the obtained compression rate is larger or smaller than the above-mentioned target value (for example, 1/100) or substantially equal to the target value. Is equal to If it is determined that the obtained compression rate is larger or smaller than the target value, the process returns to step S117.

  On the other hand, when determining that the obtained compression rate is substantially equal to the target value, the computer C1 uses the continuous component compression data and random component compression data generated in step S118 and the sample number data generated in step S112 as its own. Is output to the outside via the serial communication control unit (step S120).

  As a result of performing the processing described above, this audio data compression system converts audio data to be compressed into continuous component data having a periodicity that meets a predetermined standard, and random component data representing other components, and And entropy coding is performed separately on the continuous component data and the random component data. For this reason, in the audio data, the component caused by the voice uttered by the person and the component not caused by the entropy coding are separately encoded, and the audio data is efficiently compressed as a whole. Therefore, this voice data compression system can efficiently compress, for example, voice mails representing voices including human voices and background music.

  Further, the voice data is processed into pitch waveform data, so that the length and amplitude of a section for a unit pitch are standardized, and the influence of pitch fluctuation is removed. For this reason, the component resulting from the voice which a person utters among pitch waveform data has a strong periodicity, and this component is correctly extracted as continuous component data by the component separation part E4. Since the extracted continuous component data has a strong periodicity, entropy coding of the continuous component data is efficiently performed.

  Further, the original time length of each section of the pitch waveform data generated by the audio data compression system can be specified using the sample number data. For this reason, an external device that acquires continuous component compressed data and random component data restores the pitch waveform data using these continuous component compressed data and random component data, and then restores each section of the restored pitch waveform data. By restoring the time length to the time length in the original voice data, the original voice data can be easily restored.

The configuration of the audio data compression system is not limited to the above.
For example, the computer C1 may acquire audio data serially transmitted from the outside via a serial communication control unit. In addition, voice data may be acquired from the outside via a communication line such as a telephone line, a dedicated line, a satellite line, etc. In this case, the computer C1 includes, for example, a modem, a DSU (Data Service Unit), and the like. Just do it. Further, if the audio data is acquired from other than the recording medium drive device SMD, the computer C1 does not necessarily need to include the recording medium drive device SMD.

  The computer C1 may include a sound collection device including a microphone, an AF amplifier, a sampler, an A / D (Analog-to-Digital) converter, a PCM encoder, and the like. If the sound collection device acquires sound data by amplifying a sound signal representing sound collected by its own microphone, sampling and A / D converting, and then performing PCM modulation on the sampled sound signal Good. Note that the audio data acquired by the computer C1 is not necessarily a PCM signal.

  Further, the computer C1 writes part or all of the continuous component compressed data, the random component compressed data, and the sample number data to the recording medium set in the recording medium drive device SMD via the recording medium drive device SMD. May be. Alternatively, the data may be written in an external storage device such as a hard disk device. In these cases, the computer C1 only needs to include a control circuit such as a recording medium drive device and a hard disk controller.

  In addition, the computer C1 does not have to perform either cepstrum analysis or analysis based on the autocorrelation coefficient. In this case, the fundamental frequency obtained by one of the cepstrum analysis or the analysis based on the autocorrelation coefficient. The reciprocal of can be handled as the pitch length as it is.

Further, the amount by which the computer C1 shifts the audio data in each section of the audio data does not need to be (−Ψ). For example, the computer C1 sets δ as a real number common to each section representing the initial phase. For each section, the audio data may be phase-shifted by (−Ψ + δ). Further, the position where the computer C1 divides the audio data does not necessarily have to be the timing at which the pitch signal crosses zero, and may be the timing at which the pitch signal has a predetermined value other than 0, for example.
However, if the initial phase α is set to 0 and the audio data is divided at the timing when the pitch signal crosses zero, the value of the start point of each section becomes a value close to 0, so the audio data is divided into each section. In particular, the amount of noise included in each section is reduced.

Further, the pitch waveform data need not be interpolated by the Lagrange interpolation method, for example, the linear interpolation method may be used, or the interpolation itself may be omitted.
Further, if the fluctuation of the pitch of the audio data to be compressed is negligible, the computer C1 does not need to perform phase shift of the audio data, and the audio data is regarded as pitch waveform data and after step S113. You may make it perform the process of. Also, interpolation and resampling of audio data are not necessarily required processes.

  In step S114, the computer C1 determines whether the ratio of the amplitude of the pitch signal to the amplitude of the pitch waveform signal has reached a predetermined amount instead of determining whether the amplitude of the pitch signal has reached a predetermined amount. It may be determined. In this case, the computer C1 may move the process to step S115 when it is determined that the ratio has not reached the predetermined amount, and may move to step S116 when it is determined that the ratio has been reached.

  Further, the computer C1 does not necessarily need to entropy encode the compression characteristic data. In this case, in step S118, the computer C1 entropy encodes only the nonlinear quantized continuous component data to generate continuous component compressed data, for example. In step S120, the continuous component compression data and random component compression data generated in step S118, the sample number data generated in step S112, and the compression characteristic data generated in step S117 may be output. .

  Further, the computer C1 may omit the process of step S119. In this case, for example, in step S117, nonlinear quantized continuous component data and nonlinear quantized random component data are generated with a predetermined compression characteristic, and in step S118, Continuous component compressed data is generated by entropy encoding the nonlinear quantized continuous component data and compression characteristic data obtained in step S117, and the nonlinear quantized random component data obtained in step S117 is entropy encoded. Thus, the random component compressed data may be generated and the process may be shifted to step S120. When the computer C1 generates the nonlinear quantized continuous component data and the nonlinear quantized random component data with the predetermined compression characteristic in step S117, the computer C1 stores the compression characteristic data indicating the predetermined compression specification in advance. Alternatively, entropy coding of compression characteristic data or output to the outside may be omitted.

  In addition, instead of performing entropy coding of nonlinear quantized continuous component data in step S118, the computer C1 obtains (n + 1) pieces of continuous component data obtained in step S115 or S116 (or obtained in step S117). (N + 1) non-linear quantized continuous component data) may be linearly predictively encoded to generate continuous component compressed data. In this way, if linear predictive coding is performed on continuous component data representing components obtained by excluding random component data (components that are not considered to be caused by human-generated speech) from subband data representing speech, The data representing the voice to be uttered is accurately and efficiently subjected to linear predictive coding without being substantially affected by a component not caused by the voice uttered by a person.

  The computer C1 does not have to be a dedicated system and may be a personal computer or the like. The audio data compression program may be installed on the computer C1 from a medium (CD-ROM, MO, flexible disk, etc.) storing the audio data compression program, or the audio data compression program may be installed on a bulletin board (BBS) on a communication line. A data compression program may be uploaded and distributed via a communication line. Further, the carrier wave may be modulated with a signal representing the audio data compression program, the obtained modulated wave may be transmitted, and the apparatus that has received the modulated wave may demodulate the modulated wave to restore the audio data compression program. .

  In addition, the audio data compression program can execute the above-described processing by being activated and executed by the computer C1 under the control of the OS in the same manner as other application programs. When the OS shares a part of the above-described processing, the audio data compression program stored in the recording medium may be one that excludes the portion that controls the processing.

(Second Embodiment)
Next explained is an audio data reproducing system according to the second embodiment of the invention.
This audio data reproduction system has a physical configuration substantially the same as that of the audio data compression system shown in FIG. However, the computer C1 constituting the audio data reproduction system stores an audio data reproduction program in advance, and executes the process described later by executing the audio data reproduction program.

  Next, the operation of this audio data reproduction system will be described with reference to FIG. FIG. 5 is a diagram showing an operation flow of the audio data reproduction system.

  The user sets, for example, a recording medium on which the continuous component compressed data, random component compressed data, and sample number data in the first embodiment described above are recorded in the recording medium drive device SMD, and the computer C1 stores the audio data reproduction program. The computer C1 starts processing of the audio data reproduction program.

  Then, first, the computer C1 reads the continuous component compressed data, the random component compressed data, and the sample number data from the recording medium via the recording medium drive device SMD (FIG. 5, step S201).

  Next, the computer C1 restores (n + 1) nonlinear quantized continuous component data and compression characteristic data by decoding the read continuous component compressed data (step S202). In step S202, the computer C1 restores (n + 1) pieces of nonlinear quantized random component data by decoding the read random component compressed data.

  Next, the computer C1 compresses the instantaneous value of the waveform represented by the restored (n + 1) non-linear quantized continuous component data and the (n + 1) non-linear quantized random component data by the restored compression characteristic data. By changing according to the characteristic that is inversely transformed with the characteristic, (n + 1) continuous component data and (n + 1) random component data before nonlinear quantization are restored (step S203).

  If the compression characteristic data cannot be obtained from the continuous component compressed data in step S202, the computer C1 uses the instantaneous value of the waveform represented by the nonlinear quantized continuous component data and the nonlinear quantized random component data to obtain predetermined characteristics in step S203. The continuous component data and the random component data may be restored by changing according to the above. Alternatively, the nonlinear quantized continuous component data and the nonlinear quantized random component data may be regarded as component data and random component data, and the process may be immediately transferred from step S202 to step S204.

  Next, the computer C1 sums the instantaneous values indicated by the k-th random component data (k is an integer between 0 and n) restored in step S203 and the k-th continuous component data (however, substantially each other). A signal indicating the sum of instantaneous values at the same time is generated (step S204).

  The signal indicating the sum of the instantaneous value indicated by the kth random component data and the instantaneous value indicated by the kth continuous component data, generated in step S204, is the continuous component compressed data and random component compressed read in step S201. When the data and the sample number data are generated by, for example, the audio data compression system according to the first embodiment described above, the audio data compression system corresponds to the kth subband data generated at step S113 described above. Signal. If the kth random component data does not exist, the computer C1 may determine in step S204 that the kth continuous component is handled as it is as the kth subband data.

  Next, the computer C1 performs transformation on the total (n + 1) subband data generated in step S204 to restore pitch waveform data in which the intensity of each frequency component is represented by these subband data. (Step S205).

  The conversion performed by the computer C1 on the subband data in step S205 is a conversion that has a substantially inverse relationship to the conversion performed on the audio data in order to generate the subband data. Therefore, for example, when the subband data is generated in the above-described step S113, the computer C1 may perform inverse conversion of the conversion performed on the pitch waveform data in step S113 in step S205. Specifically, for example, if the subband data is generated by applying DCT to the pitch waveform data, the computer C1 may apply IDCT (Inverse DCT) to the subband data in step S205. Good.

Next, the computer C1 changes the time length of each section by changing the number of samples in each section of the pitch waveform data restored in step S205 to the number of samples indicated by the sample number data restored in step S202. (Step S206).
Then, the computer C1 outputs pitch waveform data in which the time length of each section is changed, that is, restored audio data (step S207).

  Note that the method in which the computer C1 outputs audio data in step S207 is arbitrary. For example, the computer C1 may output the restored audio data to the outside via its own serial communication control unit, or record it. The recording medium set in the medium drive device SMD may be written via the recording medium drive device SMD. When a control circuit such as a hard disk controller is provided, the data may be written in an external storage device such as a hard disk device. Moreover, you may make it deliver audio | voice data to the other process which computer C1 is performing itself.

  As a result of performing the above-described processing, this audio data reproduction system uses the audio data compressed by the audio data compression system of the first embodiment (or the audio data compression system of the third embodiment described later). Audio data that has been converted to the above-described continuous component compressed data, random component data, and sample number data) by any other method.

Note that the configuration of the audio data reproduction system is not limited to that described above.
For example, the computer C1 constituting the audio data reproduction system may also acquire the continuous component compressed data, random component compressed data, and sample number data serially transmitted from the outside via the serial communication control unit. In addition, continuous component compressed data, random component compressed data, and sample number data may be acquired from the outside via a communication line. In this case, the computer C1 only needs to include a modem, DSU, or the like. Further, if continuous component compressed data, random component data, and sample number data are obtained from other than the recording medium drive device SMD, the computer C1 does not necessarily have to include the recording medium drive device SMD.

  Further, the computer C1 may include an audio reproduction device including a D / A (Digital-to-Analog) converter, an AF amplifier, a speaker, and the like. In this case, in step S207, the audio reproduction device D / A converts the restored audio data to generate analog audio data, amplifies the analog audio data, and drives its own speaker. The voice represented by the voice data may be reproduced.

  Further, the computer C1 may write the restored audio data to the recording medium set in the recording medium drive device SMD via the recording medium drive device SMD. Alternatively, the data may be written in an external storage device such as a hard disk device. In these cases, the computer C1 only needs to include a control circuit such as a recording medium drive device and a hard disk controller.

  In step S206, the computer C1 adjusts the sample interval in each section of the pitch waveform data restored in step S205, thereby specifying the time length of the section from the sample number data restored in step S202. You may make it change into the length of time to be.

  Note that the computer C1 constituting the audio data reproduction system is not necessarily a dedicated system, and may be a personal computer or the like. The audio data reproduction program may be installed in the computer C1 from the medium storing the audio data compression program, or the audio data reproduction program is uploaded to a bulletin board on the communication line, and this is uploaded via the communication line. You may distribute. Further, the carrier wave may be modulated with a signal representing the audio data reproduction program, the obtained modulated wave may be transmitted, and the apparatus that has received the modulated wave may demodulate the modulated wave to restore the audio data reproduction program. . Further, the audio data reproduction program can be executed in the same manner as other application programs under the control of the OS and executed by the computer C1, thereby executing the above-described processing. When the OS shares a part of the above-described processing, the audio data reproduction program stored in the recording medium may be a program that excludes the portion that controls the processing.

(Third embodiment)
Next explained is the third embodiment of the invention.
FIG. 6 is a diagram showing the configuration of an audio data compression system according to the third embodiment of the present invention. As shown in the figure, this audio data compression system includes an audio input unit E1, a pitch waveform extraction unit E2, a subband analysis unit E3, a component separation unit E4, a data compression unit E5, and an output unit E6. Has been.

The audio input unit E1 is configured by, for example, a recording medium drive device similar to the recording medium drive device SMD in the first embodiment.
The voice input unit E1 acquires voice data representing a voice waveform by reading it from a recording medium on which the voice data is recorded, and supplies the voice data to the pitch waveform extraction unit E2. Note that the audio data has a PCM-modulated digital signal format, and represents audio sampled at a constant period sufficiently shorter than the audio pitch.

Each of the pitch waveform extraction unit E2, the subband analysis unit E3, the component separation unit E4, and the data compression unit E5 includes a processor such as a DSP or a CPU, a memory that stores a program to be executed by the processor, and the like. ing.
A single processor may perform a part or all of the functions of the pitch waveform extraction unit E2, the subband analysis unit E3, the component separation unit E4, and the data compression unit E5.

The pitch waveform extraction unit E2 divides the audio data supplied from the audio input unit E1 into sections corresponding to the unit pitch (for example, one pitch) of the audio represented by the audio data. Then, by phase-shifting and resampling each section that has been divided, the time length and phase of each section are aligned so as to be substantially the same. And the audio | voice data (pitch waveform data) by which the phase and time length of each area were arrange | equalized are supplied to the subband analysis part E3.
Further, the pitch waveform extraction unit E2 generates a pitch signal to be described later, uses the pitch signal itself as described later, and supplies the pitch signal to the component separation unit E4.
Further, the pitch waveform extraction unit E2 generates sample number data indicating the original number of samples in each section of the audio data and supplies the sample number data to the output unit E6.

  Functionally, the pitch waveform extraction unit E2 has a cepstrum analysis unit E201, an autocorrelation analysis unit E202, a weight calculation unit E203, a BPF (band pass filter) coefficient calculation unit E204, for example, as shown in FIG. , A band pass filter E205, a zero cross analysis unit E206, a waveform correlation analysis unit E207, a phase adjustment unit E208, an interpolation unit E209, and a pitch length adjustment unit E210.

  The cepstrum analysis unit E201, autocorrelation analysis unit E202, weight calculation unit E203, BPF coefficient calculation unit E204, bandpass filter E205, zero cross analysis unit E206, waveform correlation analysis unit E207, phase adjustment unit E208, interpolation unit E209, and pitch A single processor may perform a part or all of the functions of the length adjusting unit E210.

The pitch waveform extraction unit E2 specifies the length of the pitch by using both the cepstrum analysis and the analysis based on the autocorrelation function.
That is, first, the cepstrum analysis unit E201 performs cepstrum analysis on the voice data supplied from the voice input unit E1, thereby specifying the fundamental frequency of the voice represented by the voice data and generating data indicating the identified fundamental frequency. And supplied to the weight calculation unit E203.

Specifically, when audio data is supplied from the audio input unit E1, the cepstrum analysis unit E201 first converts the intensity of the audio data into a value substantially equal to the logarithm of the original value. (The base of the logarithm is arbitrary.)
Next, the cepstrum analysis unit E201 uses a fast Fourier transform technique (or other arbitrary data that generates a result of Fourier transform of discrete variables) on the spectrum of the speech data (ie, the cepstrum) whose values have been converted. This method is used.
Then, the minimum value among the frequencies giving the maximum value of the cepstrum is specified as the fundamental frequency, data indicating the identified fundamental frequency is generated and supplied to the weight calculation unit E203.

  On the other hand, when the audio data is supplied from the audio input unit E1, the autocorrelation analysis unit E202 specifies the basic frequency of the audio represented by the audio data based on the autocorrelation function of the waveform of the audio data, and specifies the specified basic Data indicating the frequency is generated and supplied to the weight calculation unit E203.

  Specifically, when the audio data is supplied from the audio input unit E1, the autocorrelation analysis unit E202 first specifies the autocorrelation function r (l) described above. Then, among the frequencies giving the maximum value of the periodogram obtained as a result of Fourier transform of the specified autocorrelation function r (l), the minimum value exceeding a predetermined lower limit value is specified as the basic frequency, and the specified basic frequency is Data shown is generated and supplied to the weight calculation unit E203.

  When a total of two pieces of data indicating the fundamental frequency are supplied one by one from the cepstrum analysis unit E201 and the autocorrelation analysis unit E202, the weight calculation unit E203 averages the absolute value of the reciprocal of the fundamental frequency indicated by these two data. Ask for. Then, data indicating the obtained value (that is, average pitch length) is generated and supplied to the BPF coefficient calculation unit E204.

  When the BPF coefficient calculation unit E204 is supplied with data indicating the average pitch length from the weight calculation unit E203 and is supplied with a zero-cross signal described later from the zero-cross analysis unit E206, the average pitch length is based on the supplied data and the zero-cross signal. And whether the zero-cross cycle is different from each other by a predetermined amount or more. When it is determined that they are not different, the frequency characteristic of the bandpass filter E205 is controlled so that the reciprocal of the zero-crossing period is the center frequency (the center frequency of the passband of the bandpass filter E205). On the other hand, when it is determined that they differ by a predetermined amount or more, the frequency characteristics of the bandpass filter E205 are controlled so that the reciprocal of the average pitch length is set as the center frequency.

The band pass filter E205 performs a function of an FIR (Finite Impulse Response) type filter having a variable center frequency.
Specifically, the bandpass filter E205 sets its own center frequency to a value according to the control of the BPF coefficient calculation unit E204. Then, the voice data supplied from the voice input unit E1 is filtered, and the filtered voice data (pitch signal) is converted into a zero-cross analysis unit E206, a waveform correlation analysis unit E207, and a continuous component extraction unit described later of the component separation unit E4. Supply to E41-0 to E41-n. The pitch signal is assumed to be digital data having a sampling interval substantially the same as the sampling interval of audio data.
The bandwidth of the bandpass filter E205 is desirably a bandwidth that always keeps the upper limit of the passband of the bandpass filter E205 within twice the fundamental frequency of the voice represented by the voice data.

The zero-cross analysis unit E206 specifies the timing when the time when the instantaneous value of the pitch signal supplied from the band-pass filter E205 becomes 0 (time when zero-crossing) comes, and the signal (zero-cross signal) indicating the specified timing is represented by the BPF coefficient. It supplies to the calculation part E204. In this way, the pitch length of the audio data is specified.
However, the zero-cross analysis unit E206 specifies the timing when the time when the instantaneous value of the pitch signal is a predetermined value other than 0 comes, and supplies a signal representing the specified timing to the BPF coefficient calculation unit E204 instead of the zero-cross signal. You may make it do.

  When the waveform correlation analysis unit E207 is supplied with the audio data from the audio input unit E1 and is supplied with the pitch signal from the bandpass filter E205, the waveform correlation analysis unit E207 receives the audio data at the timing when the boundary of the unit period (for example, 1 period) of the pitch signal comes. punctuate. Then, for each of the sections that can be divided, the correlation between the variously changed phases of the audio data in this section and the pitch signal in this section is obtained, and the phase of the audio data when the correlation becomes the highest is obtained. The phase of the audio data in this section is specified. In this way, the phase of the audio data is specified for each section.

  Specifically, for example, the waveform correlation analysis unit E207 specifies the above-described value Ψ for each section, generates data indicating the value Ψ, and represents phase data representing the phase of the audio data in the section. To the phase adjustment unit E208. Note that the time length of the section is preferably about one pitch.

  When the audio data is supplied from the audio input unit E1 and the data indicating the phase ψ of each section of the audio data is supplied from the waveform correlation analysis unit E207, the phase adjustment unit E208 sets the phase of the audio data in each interval to ( The phase of each section is aligned by shifting the phase by −Ψ). Then, the phase-shifted audio data is supplied to the interpolation unit E209.

  The interpolation unit E209 performs Lagrangian interpolation on the audio data (phase-shifted audio data) supplied from the phase adjustment unit E208, and supplies the result to the pitch length adjustment unit E210.

  When the audio data subjected to Lagrangian interpolation is supplied from the interpolation unit E209, the pitch length adjustment unit E210 substantially resamples the time lengths of the sections from each other by resampling each section of the supplied audio data. To be identical to each other. Then, the audio data (that is, pitch waveform data) in which the time lengths of the respective sections are aligned is supplied to the subband analysis unit E3.

  The pitch length adjustment unit E210 indicates the original number of samples in each section of the voice data (the number of samples in each section of the voice data at the time when the voice data is supplied from the voice input unit E1 to the pitch length adjustment unit E210). Sample number data is generated and supplied to the output unit E6.

  The subband analyzing unit E3 performs orthogonal transformation such as DCT on the pitch waveform data supplied from the pitch length adjusting unit E210, thereby obtaining a total of (n + 1) subband data consisting of this subband data. A group is generated, and this subband data group is supplied to the component separation unit E4.

  Functionally, the component separation unit E4 has (n + 1) continuous component extraction units E41-0 to E41-n and (n + 1) random component extraction units E42-0 to E42-0, for example, as shown in FIG. E42-n.

The continuous component extraction units E41-0 to E41-n perform functions of, for example, an LMS (Least Mean Square) filter or another adaptive filter (adaptive filter).
The continuous component extraction unit E41-k (k is an integer between 0 and n) determines whether or not the amplitude of the pitch signal supplied from the bandpass filter E205 has reached a predetermined amount. Then, during a period in which it is determined that the predetermined amount has not been reached, the kth subband data is filtered by filtering the kth subband data included in the subband data group supplied from the subband analysis unit E3. Among them, a component having a periodicity stronger than a certain level (hereinafter referred to as k-th continuous component data) is extracted, and the extracted k-th continuous component data is converted into a random component extraction unit E42-k and a data compression unit E5. To supply.
On the other hand, during the period when it is determined that the amplitude of the pitch signal has reached a predetermined amount, the kth subband data is used as it is as the kth continuous component data as a random component instead of filtering the kth subband data. The data is supplied to the extraction unit E42-k and the data compression unit E5.

  The random component extraction unit E42-k includes an instantaneous value indicated by the kth subband data supplied from the subband analysis unit E3 and an instantaneous value indicated by the kth continuous component data supplied from the continuous component extraction unit E42-k. A signal (hereinafter, kth random component data) indicating a difference from the value (however, a difference between instantaneous values at substantially the same time) is generated and supplied to the data compression unit E5.

The operation performed by the continuous component extraction unit E41-k is performed when the k-th subband data is determined not to ignore a component (a component such as a sound of an instrument) that is not caused by a voice uttered by a person. If a component having a certain degree of strong periodicity is extracted as the k-th continuous component data from the sub-band data and if it is determined that a component that is not caused by speech uttered by a person can be ignored, the k-th This corresponds to an operation in which the subband data is regarded as the k-th continuous component data as it is (that is, a component that is not caused by a voice uttered by a person is not present).
Therefore, when it is determined that the component that does not originate from the voice uttered by the person cannot be ignored for the kth subband data, the component of the kth subband data whose periodicity does not reach a certain level is k. In the case where it is determined that the k-th sub-band data may be considered to be composed only of the speech component uttered by the person, the strength of the k-th continuous component data is substantially 0.

  The data compression unit E5 is functionally composed of a non-linear quantization unit E51, a compression rate setting unit E52, and an entropy coding unit E53, for example, as shown in FIG.

When the non-linear quantization unit E51 is supplied with (n + 1) pieces of continuous component data from the continuous component extraction units E41-0 to E41-n, the non-linear quantization unit E51 performs non-linear compression on the instantaneous value of the waveform represented by each of these continuous component data. (N + 1) non-linear quantized continuous component data corresponding to a value obtained by quantizing a value (specifically, for example, a value obtained by substituting an instantaneous value into an upward convex function) Generate. Further, when (n + 1) pieces of random component data are supplied from the continuous component extraction units E41-0 to E41-n, they are obtained by subjecting the instantaneous values of the waveforms represented by these random component data to the nonlinear compression. (N + 1) non-linear quantized random component data corresponding to a quantized value is generated.
Then, the nonlinear quantization unit E51 supplies the generated nonlinear quantization continuous component data and nonlinear quantization random component data to the entropy coding unit E53. However, it is not necessary to generate nonlinear quantized random component data for a random component having an intensity of 0.

  The nonlinear quantization unit E51 obtains compression characteristic data for specifying the correspondence between the pre-compression value and the post-compression value of the instantaneous value from the compression rate setting unit E52, and the correspondence specified by this data. Compress according to the relationship. Specifically, for example, the nonlinear quantization unit E51 acquires data specifying the above-described function global_gain (xi) from the compression rate setting unit E52 as compression characteristic data. Then, the nonlinear quantization is performed by changing the instantaneous value of each continuous component data and each random component data after nonlinear compression to a value that is substantially equal to a value obtained by quantizing the above-described function Xri (xi). I do.

  The compression rate setting unit E52 generates the above-described compression characteristic data for specifying the correspondence (hereinafter referred to as compression characteristic) between the value before compression of the instantaneous value by the nonlinear quantization unit E51 and the value after compression. The non-linear quantization unit E51 and the entropy coding unit E53 are supplied. Specifically, for example, compression characteristic data specifying the above-described function global_gain (xi) is generated and supplied to the nonlinear quantization unit E51 and the entropy encoding unit E53.

  Note that the compression rate setting unit E52 acquires continuous component compressed data and random component compressed data, which will be described later, from the entropy encoding unit E53, for example, in order to determine the compression characteristics. Then, the continuous component compressed data and the random component compressed data acquired from the entropy encoding unit E53 with respect to the total amount of (n + 1) continuous component data and (n + 1) random component data acquired from the component separator E4. The ratio of the total amount of data is obtained, and it is determined whether or not the obtained ratio is larger than a target predetermined compression rate. If it is determined that the obtained ratio is larger than the target compression rate, the compression rate setting unit E52 determines the compression characteristics so that the compression rate is smaller than the current compression rate. On the other hand, when it is determined that the obtained ratio is equal to or less than the target compression rate, the compression characteristic is determined so that the compression rate is larger than the current compression rate.

  The entropy encoding unit E53 entropy-encodes the (n + 1) pieces of non-linear quantization continuous component data supplied from the non-linear quantization unit E51 and the compression characteristic data supplied from the compression rate setting unit E52. These pieces of data are supplied to the compression ratio setting unit E52 and the output unit E6 as continuous component compressed data. The entropy encoding unit E53 entropy-encodes (n + 1) nonlinear quantization random component data supplied from the nonlinear quantization unit E51, and these entropy-encoded data are used as random component compressed data. It supplies to the compression rate setting part E52 and the output part E6.

  The output unit E6 includes a control circuit that controls serial communication with the outside based on a standard such as USB. Note that a processor that performs some or all of the functions of the pitch waveform extraction unit E2, the subband analysis unit E3, the component separation unit E4, and the data compression unit E5 may further perform the function of the output unit E6.

  When the output unit E6 is supplied with the continuous component compression data and the random component compression data generated by the data compression unit E5 and the sample number data generated by the pitch length adjustment unit E210 of the pitch waveform extraction unit E2, the output unit E6 continuously Component compressed data, random component compressed data, and sample number data are output.

  The audio data compression system of FIG. 6 also separates audio data to be compressed into continuous component data having a periodicity that meets a predetermined standard and random component data that is other components. Each is entropy encoded separately. For this reason, in the audio data, the component caused by the voice uttered by the person and the component not caused by the entropy coding are separately encoded, and the audio data is efficiently compressed as a whole.

  Further, the voice data is processed into pitch waveform data, so that the length and amplitude of a section for a unit pitch are standardized, and the influence of pitch fluctuation is removed. For this reason, the component resulting from the voice which a person utters among pitch waveform data has a strong periodicity, and this component is correctly extracted as a continuous component by the component separation part E4. Since the extracted continuous component has a strong periodicity, entropy coding of the continuous component is performed efficiently.

  Furthermore, since the original time length of each section of the pitch waveform data can be specified using the sample number data, the time length of each section of the pitch waveform data is restored to the time length in the original audio data. The original audio data can be easily restored.

The configuration of the audio data compression system is not limited to the above.
For example, the voice input unit E1 may acquire voice data from the outside via a communication line such as a telephone line, a dedicated line, or a satellite line. In this case, the voice input unit E1 only needs to include a communication control unit including, for example, a modem or a DSU.

  The voice input unit E1 may include a sound collection device including a microphone, an AF amplifier, a sampler, an A / D converter, a PCM encoder, and the like. If the sound collection device acquires sound data by amplifying a sound signal representing sound collected by its own microphone, sampling and A / D converting, and then performing PCM modulation on the sampled sound signal Good. Note that the audio data acquired by the audio input unit E1 is not necessarily a PCM signal.

  The pitch waveform extraction unit E2 may not include the cepstrum analysis unit E201 (or autocorrelation analysis unit E202). In this case, the weight calculation unit E203 may include the cepstrum analysis unit E201 (or autocorrelation analysis unit E202). The reciprocal of the fundamental frequency obtained by (2) may be handled as the average pitch length as it is.

  The zero cross analysis unit E206 may supply the pitch signal supplied from the bandpass filter E205 as it is to the BPF coefficient calculation unit E204 as a zero cross signal.

Further, the amount by which the phase adjustment unit E208 shifts the audio data in each section of the audio data does not need to be (−Ψ), and the position where the waveform correlation analysis unit E207 divides the audio data is not necessarily a pitch signal. Need not be at the timing of zero crossing.
Further, the interpolation unit E209 does not necessarily perform the phase-shifted audio data interpolation by the Lagrangian interpolation method. For example, the interpolation unit E209 may omit the interpolation unit E209 and the phase adjustment unit E208 The data may be immediately supplied to the pitch length adjustment unit E210.

Further, the output unit E6 may output to the outside via phoneme data, sample number data, a communication line, or the like. When outputting data via a communication line, the output unit E6 only needs to include a communication control unit such as a modem or a DSU.
Further, the output unit E6 may include a recording medium drive device. In this case, the output unit E6 sets continuous component compressed data, random component compressed data, and sample number data in this recording medium drive device. You may make it write in the storage area of a recording medium.
Note that a single modem, DSU, or recording medium drive device may constitute the audio input unit E1 and the output unit E6.

  Further, instead of determining whether or not the amplitude of the pitch signal has reached a predetermined amount, the continuous component extraction unit E41-k determines whether or not the ratio of the amplitude of the pitch signal to the amplitude of the pitch waveform signal has reached a predetermined amount. It may be determined.

  Further, the entropy encoding unit E53 does not necessarily need to entropy encode the compression characteristic data. In this case, for example, the compression rate setting unit E52 converts the compression characteristic data generated by itself into the nonlinear quantization unit E51 and the output unit. The output unit E6 may output the compression characteristic data supplied from the compression rate setting unit E52 in addition to the continuous component compression data, the random component compression data, and the sample number data.

  The data compression unit E5 does not necessarily include the compression rate setting unit E52. In this case, for example, the nonlinear quantization unit E51 has nonlinear compression continuous component data and nonlinear quantization random component data with predetermined compression characteristics. The entropy encoding unit E53 entropy-encodes the continuous component compressed data by entropy encoding the compression characteristic data indicating the predetermined compression characteristic and the non-linear quantization continuous component data generated by the non-linear quantization unit E51. It may be generated. The entropy encoding unit E53 may omit the entropy encoding of the compression characteristic data.

  Further, the entropy coding unit E53 generates continuous component compressed data by performing linear predictive coding on continuous component data or nonlinear quantized continuous component data instead of entropy coding of nonlinear quantized continuous component data. It may be.

(Fourth embodiment)
Next, a fourth embodiment of the present invention will be described by taking an audio data reproduction system as an example.
As shown in FIG. 10, the audio data reproduction system includes a data input unit D1, an entropy code decoding unit D2, a nonlinear inverse quantization unit D3, a component combination unit D4, a subband synthesis unit D5, The data restoration unit D6 and the speech synthesis unit D7 are configured.

  The data input unit D1, entropy code decoding unit D2, nonlinear inverse quantization unit D3, component combination unit D4, subband synthesis unit D5 and speech data restoration unit D6 are all processors such as DSPs and CPUs, Is configured by a memory for storing a program to be executed. In addition, a single processor functions part or all of the data input unit D1, the entropy code decoding unit D2, the nonlinear inverse quantization unit D3, the component combination unit D4, the subband synthesis unit D5, and the speech data restoration unit D6. You may make it perform.

  The data input unit D1 acquires the above-described continuous component compressed data, random component compressed data, and sample number data from the outside, and among these acquired data, the continuous component compressed data and the random component compressed data are entropy code decoding units. The sample number data is supplied to D2, and the sample number data is supplied to the audio data restoration unit D6.

  The data input unit D1 can use any method for acquiring the continuous component compressed data, the random component compressed data, and the sample number data. For example, the data input unit D1 acquires the compressed phoneme data recorded on the computer-readable recording medium. Alternatively, it may be acquired by receiving these data serially transmitted in a mode compliant with standards such as Ethernet (registered trademark), USB, IEEE1394, or RS232C, or these data transmitted in parallel. . The data input unit D1 may acquire these data stored in an external server by a technique such as downloading via a network such as the Internet.

  The data input unit D1 further includes, for example, a recording medium drive device that reads data from a recording medium in accordance with instructions from a processor or the like when reading continuous component compressed data, random component compressed data, or sample number data from a recording medium. It only has to have. Further, when receiving these serially transmitted data, it is only necessary to further include a control circuit for controlling serial communication with the outside in accordance with standards such as Ethernet (registered trademark), USB, IEEE 1394, or RS232C. .

  The entropy code decoding unit D2 restores (n + 1) non-linear quantized continuous component data and compression characteristic data by decoding the continuous component compressed data supplied from the data input unit D1. Then, these restored data are supplied to the non-linear inverse quantization unit D3. The entropy code decoding unit D2 restores (n + 1) non-linear quantization random component data by decoding the random component compressed data supplied from the data input unit D1, and restores the restored non-linear quantization Random component data is also supplied to the nonlinear inverse quantization unit D3.

  When the non-linear inverse quantization unit D3 is supplied with (n + 1) non-linear quantization continuous component data, (n + 1) non-linear quantization random component data and compression characteristic data from the entropy code decoding unit D2, By changing the instantaneous value of the waveform represented by the non-linear quantized continuous component data and the non-linear quantized random component data according to the characteristic inversely transformed from the compression characteristic indicated by the compression characteristic data, (N + 1) continuous component data and (n + 1) random component data are restored. Then, the restored continuous component data and random component data are supplied to the component combination unit D4.

  When the entropy code decoding unit D2 cannot obtain the compression characteristic data from the continuous component compressed data, the nonlinear inverse quantization unit D3 instantiates the waveform represented by the nonlinear quantized continuous component data and the nonlinear quantized random component data. Continuous component data and random component data may be restored by changing the value according to a predetermined characteristic, or nonlinear quantized continuous component data and nonlinear quantized random component data are regarded as continuous component data and random component data. Then, it may be supplied to the component coupling part D4 as it is.

  When the component combination unit D4 is supplied with (n + 1) continuous component data and (n + 1) random component data from the nonlinear inverse quantization unit D3, the kth random component supplied from the nonlinear inverse quantization unit D3. A signal indicating the sum of the instantaneous values indicated by the component data and the kth continuous component data (however, the sum of the instantaneous values at substantially the same time) is generated and supplied to the subband synthesizing unit D5. . This signal indicating the sum of the instantaneous value indicated by the kth random component data and the instantaneous value indicated by the kth continuous component data corresponds to the kth subband data generated by the subband analysis unit E3. Signal. If the kth random component data does not exist, the component combination unit D4 may treat the kth continuous component data as it is as the kth subband data.

  When the subband synthesizing unit D5 is supplied with a total of (n + 1) subband data from the component combining unit D4, the subband synthesizing unit D5 performs conversion on these subband data so that the intensity of each frequency component is obtained by the subband data. Is restored, and the restored pitch waveform data is supplied to the audio data restoration unit D6.

  The conversion performed on the subband data by the subband synthesizing unit D5 is a conversion that has a substantially inverse relationship to the conversion performed on the audio data in order to generate the subband data. Therefore, for example, when the subband data is generated by the above-described subband analysis unit E3 (or the processing in step S113 described above), the subband synthesis unit D5 performs the subband analysis unit E3 (or the above described above). The inverse conversion of the conversion performed in step S113) may be performed. Specifically, for example, when the subband data is generated by applying DCT to phonemes, the subband synthesizing unit D5 may apply IDCT (Inverse DCT) to the subband data.

The audio data restoration unit D6 adjusts the number of samples or the sample interval in each section of the pitch waveform data supplied from the subband synthesis unit D5, and the time length of the section is supplied from the data input unit D1. The time length specified by the sample number data is set.
Then, the voice data restoration unit D6 outputs the pitch waveform data in which the time length of each section is changed, that is, the restored voice data.

  The method of outputting the audio data by the audio data restoration unit D6 is arbitrary. For example, the audio represented by the audio data is reproduced via a D / A (Digital-to-Analog) converter or a speaker (not shown). You may do it. Further, it may be sent to an external device or a network via an interface circuit (not shown), or may be written to a recording medium set in a recording medium drive device (not shown) via this recording medium drive device. Further, the processor performing the function of the voice data restoration unit D6 may deliver the voice data to another process being executed by itself.

It is a block diagram which shows the structure of the audio | voice data compression system which concerns on 1st Embodiment of this invention. It is a figure which shows the first half of the flow of operation | movement of the audio | voice data compression system of FIG. It is a figure which shows the second half of the flow of operation | movement of the audio | voice data compression system of FIG. (A) And (b) is a graph which shows the waveform of the audio | voice data before phase-shifting, (c) is a graph showing the waveform of the audio | voice data after phase-shifting. It is a figure which shows the first half of the flow of operation | movement of the audio | voice data reproduction | regeneration system based on 2nd Embodiment of this invention. It is a block diagram which shows the structure of the audio | voice data compression system which concerns on 3rd Embodiment of this invention. It is a block diagram which shows the structure of the pitch waveform extraction part of the audio | voice data compression system of FIG. It is a block diagram which shows the structure of the component separation part of the audio | voice data compression system of FIG. It is a block diagram which shows the structure of the data compression part of the audio | voice data compression system of FIG. It is a block diagram which shows the structure of the audio | voice data reproduction | regeneration system which concerns on 4th Embodiment of this invention.

Explanation of symbols

C1 computer SMD recording medium drive device E1 voice input unit E2 pitch waveform extraction unit E201 cepstrum analysis unit E202 autocorrelation analysis unit E203 weight calculation unit E204 BPF coefficient calculation unit E205 band pass filter E206 zero cross analysis unit E207 waveform correlation analysis unit E208 phase adjustment Unit E209 interpolation unit E210 pitch length adjustment unit E3 subband analysis unit E4 component separation unit E41-0 to E41-n continuous component extraction unit E42-0 to E42-n random component extraction unit E5 data compression unit E51 nonlinear quantization unit E52 Compression rate setting unit E53 Entropy encoding unit E6 Output unit D1 Data input unit D2 Entropy code decoding unit D3 Nonlinear inverse quantization unit D4 Component combination unit D5 Subband synthesis unit D6 Phoneme data restoration unit D7 Speech synthesis unit

Claims (12)

Subband extraction means for generating a subband signal representing a temporal change in the intensity of the fundamental frequency component and the harmonic component of the audio signal to be compressed representing the waveform of the audio;
When it is determined that the amplitude of the pitch signal obtained by filtering the audio signal does not reach a predetermined amount, the subband signal has a continuous component having a periodicity that meets a predetermined criterion, and A random component corresponding to the subband signal excluding the continuous component is separated, and the subband signal is treated as the continuous component when it is determined that the amplitude of the pitch signal has reached the predetermined amount. Component separation means;
Encoding means for performing entropy encoding or linear predictive encoding on the continuous components ;
Audio signal compression apparatus characterized by a Turkey provided with. By acquiring the audio signal to be compressed representing the waveform of the audio, and dividing the audio signal into a plurality of intervals corresponding to the unit pitch of the audio, the phases of these intervals are made substantially the same, Audio signal processing means for processing the signal into a pitch waveform signal;
Subband extraction means for generating a subband signal representing a temporal change in intensity of the fundamental frequency component and the harmonic component of the pitch waveform signal;
When it is determined that the amplitude of the pitch signal obtained by filtering the audio signal does not reach a predetermined amount, the subband signal has a continuous component having a periodicity that meets a predetermined criterion, and A random component corresponding to the subband signal excluding the continuous component is separated, and the subband signal is treated as the continuous component when it is determined that the amplitude of the pitch signal has reached the predetermined amount. Component separation means;
Encoding means for performing entropy encoding or linear predictive encoding on the continuous components ;
Audio signal compression apparatus characterized by a Turkey provided with. The encoding means performs entropy encoding on the result of nonlinear quantization of the continuous component and / or the result of nonlinear quantization of the random component.
Characterized in that, the speech signal compression apparatus according to claim 1 or 2. The encoding means generates data indicating a quantization characteristic of the nonlinear quantization;
Characterized in that, the speech signal compression apparatus according to claim 3. The encoding means determines a quantization characteristic of the nonlinear quantization based on a data amount of a continuous component and / or a random component that have been entropy-encoded in the past, and matches the determined quantization characteristic. Perform nonlinear quantization,
Characterized in that, the speech signal compression apparatus according to claim 3 or 4. An entropy code is extracted by extracting continuous components having a periodicity that meets a predetermined standard from subband signals representing temporal changes in the fundamental frequency components and harmonic component intensities of audio signals to be compressed that represent speech waveforms. Decoding means for obtaining an input signal corresponding to a signal that has been subjected to normalization or linear predictive coding, and restoring the continuous component by decoding the input signal;
If the random component corresponding to the restored continuous component minus from said sub-band signal is present, by obtains the random component, adds the random component to the connected component, the sub-band A subband signal restoration unit that restores a signal and treats the restored continuous component as the subband signal when the random component does not exist ;
And voice signal restoring means for restoring the audio signal of the compressed object based on the reconstructed subband signals,
Speech signal decompression apparatus according to claim and Turkey provided with. An audio signal compression method executed by an audio signal compression apparatus having a processor,
The processor generates a subband signal representing a temporal change in intensity of a fundamental frequency component and a harmonic component of an audio signal to be compressed representing an audio waveform;
When the processor determines that the amplitude of the pitch signal obtained by filtering the audio signal does not reach a predetermined amount, the processor has a periodicity that meets a predetermined criterion from the subband signal. When the component and a random component corresponding to the sub-band signal obtained by removing the continuous component are separated and it is determined that the amplitude of the pitch signal has reached the predetermined amount, the sub-band signal is Treat as a continuous component,
The processor performs entropy coding or linear predictive coding on the continuous components;
An audio signal compression method. An audio signal compression method executed by an audio signal compression apparatus having a processor,
When the processor acquires an audio signal to be compressed representing a waveform of an audio and divides the audio signal into a plurality of intervals corresponding to a unit pitch of the audio, the phases of these intervals are made substantially the same. By processing the audio signal into a pitch waveform signal,
The processor generates a subband signal representing a temporal change in intensity of a fundamental frequency component and a harmonic component of the pitch waveform signal;
When the processor determines that the amplitude of the pitch signal obtained by filtering the audio signal does not reach a predetermined amount, the processor has a periodicity that meets a predetermined criterion from the subband signal. When the component and the random component corresponding to the sub-band signal obtained by removing the continuous component are separated and it is determined that the amplitude of the pitch signal has reached a predetermined amount, the sub-band signal is Treat as an ingredient,
The processor performs entropy coding or linear predictive coding on the continuous components;
An audio signal compression method. An audio signal restoration method executed by an audio signal restoration device having a processor,
The processor extracts a continuous component having a periodicity enough to meet a predetermined standard from subband signals representing temporal changes in intensity of fundamental frequency components and harmonic components of a compression target speech signal representing a speech waveform. To obtain an input signal corresponding to the one subjected to entropy coding or linear prediction coding, and to restore the continuous component by decoding the input signal,
Wherein the processor, when the random component corresponding to the restored continuous component minus from said sub-band signal is present, acquires the random component, by adding the random component to the continuous component The subband signal is restored, and when the random component is not present, the restored continuous component is treated as the subband signal,
The processor restores the audio signal to be compressed based on the restored subband signal;
A method for restoring an audio signal. Computer
Subband extraction means for generating a subband signal representing a temporal change in intensity of a fundamental frequency component and a harmonic component of an audio signal to be compressed representing a waveform of the audio ;
When it is determined that the amplitude of the pitch signal obtained by filtering the audio signal does not reach a predetermined amount, the subband signal has a continuous component having a periodicity that meets a predetermined criterion, and A random component corresponding to the subband signal excluding the continuous component is separated, and the subband signal is treated as the continuous component when it is determined that the amplitude of the pitch signal has reached the predetermined amount. Component separation means ,
Encoding means for performing entropy encoding or linear predictive encoding on the continuous components ;
Program to function as a. Computer
By acquiring the audio signal to be compressed representing the waveform of the audio, and dividing the audio signal into a plurality of intervals corresponding to the unit pitch of the audio, the phases of these intervals are made substantially the same, Audio signal processing means for processing a signal into a pitch waveform signal ;
Subband extraction means for generating a subband signal representing a temporal change in intensity of the fundamental frequency component and the harmonic component of the pitch waveform signal ;
When it is determined that the amplitude of the pitch signal obtained by filtering the audio signal does not reach a predetermined amount, the subband signal has a continuous component having a periodicity that meets a predetermined criterion, and A random component corresponding to the subband signal excluding the continuous component is separated, and the subband signal is treated as the continuous component when it is determined that the amplitude of the pitch signal has reached the predetermined amount. Component separation means ,
Encoding means for performing entropy encoding or linear predictive encoding on the continuous components ;
Program to function as a. Computer
An entropy code is extracted by extracting continuous components having a periodicity that meets a predetermined standard from subband signals representing temporal changes in the fundamental frequency components and harmonic component intensities of audio signals to be compressed that represent speech waveforms. A decoding unit that obtains an input signal corresponding to a signal that has been subjected to normalization or linear predictive coding, and restores the continuous component by decoding the input signal ;
If the random component corresponding to the restored continuous component minus from said sub-band signal is present, by obtains the random component, adds the random component to the connected component, the sub-band A subband signal restoration unit that restores a signal and treats the restored continuous component as the subband signal when the random component is not present ;
Voice signal restoring means for restoring the audio signal of the compressed object based on the reconstructed subband signals,
Program to function as a.

Images