KR100452955B1 - Voice encoding method, voice decoding method, voice encoding device, voice decoding device, telephone device, pitch conversion method and medium - Google Patents

Abstract

In the case where an audio signal is encoded or decoded, pitch control can be performed by simple processing and structure. When the speech signal performs sinusoidal encoding on each encoder obtained by dividing the speech signal on the time axis by a predetermined encoder, a linear predictive residual of the speech signal is obtained, and the resulting speech coded data is processed and the sinusoidal wave is processed. The pitch component of speech encoded data encoded by the analysis encoding is changed by a predetermined calculation process in the pitch conversion unit.

Description

Voice encoding method, voice decoding method, voice encoding device, voice decoding device, telephone device, pitch conversion method and medium

The present invention provides an encoding method and a decoding method applied when a speech signal is highly efficient encoded or decoded, an encoding device, a decoding device and a telephone device to which an encoding method and a decoding method are applied, and processing data of encoding and decoding are recorded. It is about various media.

Various encoding methods are known in which signal compression is performed by utilizing the statistical properties of audio signals (where audio signals include voice signals and sound signals) and human hearing properties in the time domain and the frequency domain. Coding methods are broadly classified into coding in the time domain, coding in the frequency domain, and analysis coding.

Examples of high efficiency coding such as voice signals include MBE (multiband excitation) coding, SBE (singleband excitation) or sine wave synthesis coding, harmonic coding, SBC (sub-band coding) Split coding), LPC (linear predictive coding), DCT (discrete cosine transform), MDCT (modified DCT), FFT (fast free-switched transform), and the like are known.

When a speech signal is encoded using the various encoding methods or when the encoded speech signal is decoded, it is required to change the pitch of the speech without changing the phoneme of the speech.

In the conventional high efficiency encoding apparatus and high efficiency decoding apparatus for speech signals, it is necessary to connect a separate pitch control apparatus and perform pitch conversion without considering the pitch change, resulting in a disadvantage of a complicated structure.

In view of the foregoing, it is an object of the present invention to accurately execute desired pitch control with a simple process and structure without changing the phoneme when performing the encoding and decoding processing of a speech signal.

In order to solve the above-mentioned problem, a predetermined encoding apparatus divides a speech signal on the time axis, obtains linear predictive residuals in each encoder, and performs sinusoidal encoded encoding on the linear predicted residuals, which are encoded by sinusoidal encoded encoding. The pitch component of speech encoded data is changed by a predetermined calculation process in accordance with the present invention.

According to the present invention, the pitch conversion can be simply performed without changing the phoneme components in the calculation processing of the speech encoded data encoded by the sinusoidal analysis encoding.

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

1 is a block diagram showing a basic configuration of an example of an audio encoding device, and FIG. 3 is a block diagram showing a detailed configuration thereof.

The basic concept of speech processing according to the embodiment of the present invention will be described below. On the encoding side of the audio signal, a technique or number of dimensional transformations of data transformation previously proposed by the present invention and described in Japanese Laid-Open No. 6-51800 is used. In the quantization of the amplitude of the spectral envelope using the technique, vector quantization is performed with a number of harmonics maintained at a certain number, i.e., a certain number of dimensions. Since the shape of the spectral envelope does not change, the phoneme components included in the speech component do not change.

In the basic concept, the speech signal encoding apparatus of FIG. 1 obtains short-term prediction residuals such as LPC (linear predictive encoding) residuals, and performs a phase transmission on an input audio signal and a first encoding unit 110 that performs sine wave analysis encoding such as harmonic encoding. And a second encoding unit 120 which encodes by waveform encoding. The first encoder 110 is used to encode the V (voiced sound) part of the input signal, while the second encoder 120 is used to encode the UV (unvoiced) part of the input signal.

In the first encoding unit 110, a structure for performing sinusoidal analysis encoding such as, for example, harmonic encoding or multiband excitation (MBE) encoding in the LPC residual is used. In the second encoding unit 120, a structure of code excitation linear prediction (CELP) encoding is used, for example, by vector quantization as a closed loop search of an optimal vector using a method of synthesis.

In the example of FIG. 1, the audio signal supplied to the input terminal 101 is sent to the LPC analysis and quantization unit 113 of the LPC inverse filter 111 and the first encoding unit 110. The LPC coefficient or so-called α parameter obtained by the LPC analysis and quantization unit 113 is sent to the LPC inverse filter 111. By the LPC inverse filter 111, a linear prediction residual (LPC prediction) of the input audio signal is obtained. In the LPC analysis and quantization unit 113, the quantized output of the LSP (Linear Spectrum Pair) is obtained as described below and sent to the output terminal 102. In the LPC inverse filter 111, the LPC residual is sent to the sinusoidal analysis coding unit 114.

In the sinusoidal analysis coding unit 114, pitch detection and spectral envelope amplitude calculation are performed. In addition, the V (voiced sound) / UV (unvoiced) determination is performed by the V / UV determiner 115. The spectrum envelope amplitude data from the sine wave analysis coding unit 114 is sent to the vector quantization unit 116. As a vector quantization output of the spectral envelope, the codebook index is sent from the vector quantization unit 116 to the output terminal 103 via the switch 117. Pitch data, which is the pitch component supplied from the sine wave analysis encoding unit 114, is sent to the output terminal 104 via the pitch conversion unit 119 and the switch 118. The V / UV determination output is sent from the V / UV determination unit 115 to the output terminal 105, and is sent to the switches 117 and 118 as the control signal. In the above-mentioned voiced sound V, the above-described index and pitch are selected and obtained at the output terminals 103 and 104, respectively.

Upon receiving the pitch conversion command, the pitch conversion unit 119 changes the pitch data and performs the pitch conversion by calculation processing based on the command. The detailed processing will be described below.

In the vector quantization in the vector quantization unit 116, for example, dummy data for interpolating values from the last data in the block to the first data in the block with respect to amplitude data of one block of the effective band on the frequency axis, Alternatively, the last data and dummy data extending the first data are added appropriately at the end and the beginning. Increase the number of data to N _F. Next, the amplitude data of the number of times O _S is obtained by performing oversampling of O _S times (for example, 8 times) of the band limit type. The amplitude data ((m _MX +1) XO _S ) of this O _S times is linearly interpolated and expanded to more N _M (for example, 2048) data. The N _M pieces of data are extracted and converted into the predetermined number M (for example, 44 pieces) of data, and then vector quantized.

In this example, the second encoding unit 120 has a CELP (signal excitation linear prediction) encoding structure, and the weighted speech obtained by synthesizing the output from the noise encoding field 121 by the weighted synthesis filter 122 is obtained. Send to the subtractor 123, take out the error from the audio signal supplied to the input terminal 101 via the audio weight filter 125, send this error to the distance calculating circuit 124 to calculate the distance, The vector quantization of time-axis waveforms using closed-loop search using the "Analysis by Synthesis" method, which searches for the vector with the minimum error in the noise coding field 121, is performed. This CELP encoding is used for encoding the unvoiced portion as described above, and the codebook index as the UV data from the noise coding field 121 has the result of the V / UV determination from the V / UV determination 115. It is taken out of the output terminal 107 via the switch 127 which is turned on when it is silent (UV).

Next, a basic configuration of a speech signal decoding apparatus for decoding speech encoded data encoded by the speech signal encoding apparatus of FIG. 1 will be described with reference to FIG.

In FIG. 2, the codebook index as the quantized output of the LSP (line spectrum pair) in the output terminal 102 described in FIG. 1 is input to the input terminal 202. As shown in FIG. Inputs 203, 204, and 205 are input to respective outputs from the output terminals 103, 104, and 105 of FIG. 1, that is, an index, pitch, and V / UV determination output as envelope quantization output, respectively. In addition, an index as data for UV (silent sound) from the output terminal 107 of FIG. 1 is input to the input terminal 207.

The index of the LPC residual from the input terminal 203 as the spectral envelope quantization output is sent to the inverse vector quantizer 212, inverse vector quantized, and sent to the data converter 270. In addition, pitch data at the input terminal 204 is supplied to the data converter 270 via the pitch data converter 215. In the data converter 270, the amplitude data of the number corresponding to the set pitch of the spectral envelope of the LPC residual and the changed pitch data are sent to the voiced sound synthesis unit 211. Here, in the pitch converting section 215, when the instruction to perform the pitch conversion is made, the pitch data can be changed and the pitch transform can be performed by the calculation processing based on the instruction. The detailed processing will be described later.

The voiced sound synthesis unit 211 synthesizes the LPC (linear predictive encoding) residual of the voiced sound portion by sine wave synthesis. The voiced sound synthesis section 211 is supplied with the V / UV judgment output from the terminal 205. The LPC residual of the voiced sound from the voiced sound synthesis unit 211 is sent to the LPC synthesis filter 214. The index of the UV data from the input terminal 207 is sent to the unvoiced speech synthesis section 220, and the LPC residual of the unvoiced speech section is taken out by referring to the noise code field. This LPC residual is also sent to the LPC synthesis filter 214. In the LPC synthesis filter 214, the LPC residual of the voiced sound portion and the LPC residual of the unvoiced sound portion are each independently performed. Alternatively, the LPC synthesis process may be performed on the addition of the LPC residual of the voiced sound portion and the LPC residual of the unvoiced sound portion. Here, the index of the LSP from the input terminal 202 is sent to the LPC parameter regeneration unit 213 to take out the α parameter of the LPC, which is sent to the LPC synthesis filter 214. The audio signal obtained by LPC synthesis by the LPC synthesis filter 214 is taken out from the output terminal 201.

Next, a more basic configuration of the audio signal encoding apparatus shown in FIG. 1 will be described with reference to FIG. 3. 3, the same code | symbol is attached | subjected to the part corresponding to FIG.

In the audio signal encoding apparatus shown in Fig. 3, the audio signal supplied to the input terminal 101 is subjected to a filter process of removing unnecessary band signals by the high pass filter (HPF) 109. Then, it is sent to the LPC analysis circuit 132 and the LPC inverse filter circuit 111 of the LPC (Linear Prediction Encoding) analysis and quantization unit 113.

The LPC analysis circuit 132 of the LPC analysis and quantization unit 113 calculates the linear predictive coefficient, so-called α parameter, by autocorrelation through a Hamming window with a length of about 256 samples of the input signal waveform as one block. The interval between framing, which is the unit of data output, is about 160 samples. When the sampling frequency fs is 8 kHz, for example, one frame interval is 20 msec with 160 samples.

The? Parameter from the LPC analysis circuit 132 is sent to the? → LSP conversion circuit 133 and converted into a line spectrum pair (LSP) parameter. This converts the α parameter obtained as a direct filter coefficient into, for example, ten or five pairs of LSP parameters. The conversion is performed using, for example, the Newton Labson method. The conversion to this LSP parameter is because the interpolation characteristic is superior to the? Parameter.

The LSP parameter from the? -LSP conversion circuit 133 is matrix- or vector-quantized by the LSP quantizer 134. At this time, vector quantization may be performed by taking the difference between the frames, and matrix quantization may be performed by collecting a plurality of frames. Here, 20 msec is used as one frame, and LSP parameters calculated every 20 msec are collected for two frames to perform matrix quantization and vector quantization.

The quantization output from the LSP quantizer 134, that is, the index of the LSP quantization, is taken out via the terminal 102. The quantized LSP vector is sent to the LSP interpolation circuit 136.

The LSP interpolation circuit 136 interpolates the vector of the quantized LSP every 20 msec or 40 msec, and makes the rate 8 times. That is, the LSP vector is updated every 2.5 msec. When the residual waveform is analyzed and synthesized by the harmonic coded decoding method, the envelope of the synthesized waveform is considerably smoothly inclined and becomes a smooth waveform. Therefore, if the number of LPS changes abruptly every 20 msec, a noise may occur. By gradually changing the LPC coefficient every 2.5 msec, occurrence of such anomalies can be prevented.

In order to perform inverse filtering of the input speech using the LPS vector every 2.5 msec in which such interpolation has been performed, the LSP-? Convert to α parameter. The output from the LSP? Alpha conversion circuit 137 is sent to the LPC inverse filter circuit 111. In this LPC inverse filter 111, the reverse filtering process is performed by the α parameter updated every 2.5 msec to obtain a smooth output. The output from the LPC inverse filter 111 is sent to a sinusoidal analysis encoder 114, specifically, an orthogonal transform circuit 145 of a harmonic encoding circuit, for example a DFT (discrete Fourier transform) circuit.

The α parameter from the LPC analysis circuit 132 of the LPC analysis and quantization unit 113 is sent to the auditory weighting filter calculation circuit 139 to obtain data for auditory weighting. The vector quantizer 116 and the auditory weighting filter 125 and the auditory weighting synthesis filter 122 of the second encoder 120 are sent.

A sine wave analysis coding unit 114 such as a harmonic encoding circuit analyzes the output from the LPC inverse filter 111 by the method of harmonic encoding. That is, pitch detection, calculation of the amplitude Am of each harmonics, discrimination of voiced sound (V) / unvoiced sound (UV), and dimensional conversion of the number of harmonic envelopes or amplitudes Am varying by pitch I am numbered.

In the specific example of the sinusoidal analysis coding unit 114 shown in FIG. 3, general harmonic encoding is assumed. In particular, in the case of MBE (Multiband Excitation) encoding, it is modeled on the assumption that voiced and unvoiced portions exist for each so-called band in the frequency axis region of the simultaneous angle (in the same block or frame). do. In other harmonic encoding, an alternative determination is made as to whether voice in one block or frame is voiced or unvoiced. In addition, "UV to a frame" in the following description means that the whole band is UV when applied to MBE encoding.

In the open loop pitch search unit 141 of the sine wave analysis encoder 114 of FIG. 3, the input audio signal from the input terminal 101 is input, and the zero crossing counter 142 is the HPF (high pass filter) 109. The signals from) are supplied respectively. The LPC residual or linear prediction residual from the LPC inverse filter 111 is supplied to the orthogonal transform circuit 145 of the sine wave analysis encoding unit 114. In the open loop pitch search unit 141, the LPC residual of the input signal is taken to perform a relatively rough pitch search by the open loop, and the extracted rough pitch data is sent to the high precision pitch search 146 to be described later. High-precision pitch search (precision search of pitch) is performed. In addition, in the open loop pitch search unit 141, the normalized autocorrelation maximum value r (P) obtained by normalizing the maximum value of the autocorrelation of the LPC residual with power is taken out together with the rough pitch data, and the V / UV ( Voiced sound / unvoiced sound).

In the orthogonal transform circuit 145, an orthogonal transform process such as a DFT (discrete Fourier transform) is performed, and the LPC residual on the time axis is converted into spectral amplitude data on the frequency axis. The output from the orthogonal transformation circuit 145 is sent to the high precision pitch search unit 146 and the spectrum evaluation unit 148 for evaluating the spectral amplitude or envelope.

The fine pitch search unit 146 includes relatively rough coarse pitch data extracted from the open loop pitch search unit 141 and a frequency axis that is, for example, DFTed by the orthogonal transform unit 145. The data is supplied. In this high-precision pitch search section 146, the sample is shaken by ± several samples at the 0.2 to 0.5 hour centering on the rough pitch data value, and is driven into the optimum value of the fine pitch data of the floating point portion (floating). As a method of fine search at this time, a so-called synthesis analysis method is used, and the pitch is selected so that the synthesized power spectrum is closest to the power spectrum of the original sound. The pitch data from the high precision pitch search unit 146 by the closed loop is sent to the output terminal 104 via the switch 118. And when pitch conversion is needed, pitch conversion can be performed by the process mentioned later in the pitch conversion part 119. FIG.

In the spectrum evaluation unit 148, the spectral envelope, which is the size and set of each harmonic, is evaluated based on the spectral amplitude and pitch as the orthogonal transform output of the LPC residual, and the high precision pitch search unit 146 and the V / UV (Voiced / unvoiced) is sent to the judging unit 115 and the auditory weighting vector quantizer 116.

The V / UV (voiced / unvoiced) determination unit 115 outputs from the quadrature conversion circuit 145, the optimum pitch from the high precision pitch search unit 146, and the spectral amplitude from the spectrum evaluation unit 148. The V / UV determination of the frame is based on the data, the normalized autocorrelation maximum value r (P) from the open loop pitch search unit 141, and the zero cross count value from the zero clock counter 142. Is done. The boundary position of the V / UV determination result for each band in the case of MBE may also be a condition of the V / UV determination of the frame. The determination output from this V / UV determination section 115 is taken out via the output terminal 105.

By the way, at the output of the spectrum evaluation unit 148 or at the input of the vector quantizer 116, a data number conversion (a kind of sampling rate conversion) is provided. This data number converter is intended to make a certain number of amplitude data (| Am |) of the envelope in consideration of the fact that the number of division bands on the high frequency axis is different and the number of data is different depending on the pitch. That is, for example, when the effective band is up to 3400 kHz, the effective band is divided into 8 bands to 63 bands according to the pitch, and thus the number (m _MX ) of the amplitude data (| Am |) obtained for each of these bands. +1) also changes. For this reason, the data number converting section 119 converts this variable number (m _MX +1) amplitude data into a constant number M pieces, for example, 44 pieces of data.

The constant number M pieces (for example, 44 pieces) of amplitude data or envelope data from the data number conversion section provided in the output section of the spectrum evaluation section 148 or the input section of the vector quantizer 116 are vector quantizers. By 116, a predetermined number, for example, 44 pieces of data, are gathered into a vector and weighted vector quantization is performed. This weighting is given by the output from the auditory weighting filter calculation circuit 139. The index of the envelope from the vector quantizer 116 is taken out of the output terminal 103 via the switch 117. In addition, prior to the weight vector quantization, a difference between frames using an appropriate leak coefficient may be taken for a vector formed from a predetermined number of data.

Next, the second encoding unit 120 will be described. The second encoding unit 120 has a so-called CELP (signal excitation linear prediction) encoding structure, and is particularly used for encoding unvoiced portions of input audio signals. In the CELP encoding configuration for the unvoiced portion, a noise output corresponding to the LPC residual of the unvoiced sound, which is a representative value output from a noise code book, a so-called stochastic code book 121, is audited through the gain circuit 126. It is sent to the weighted synthesis filter 122. The weighted synthesis filter 122 performs LPC synthesis processing on the input noise, and sends the obtained weighted unvoiced sound signal to the subtractor 123. The subtractor 123 inputs an audio weighted signal from the audio weight filter 125 to the audio signal supplied from the input terminal 101 through the HPF (high pass filter) 109. The difference or error with the signal is taken out. The error is sent to the distance calculating circuit 124 to calculate the distance, and the noise value encoding field 121 searches for the representative value vector so that the error is minimized. The vector quantization of time-axis waveforms using closed loop search using such an analytical method is performed.

As data for the UV (unvoiced) portion from the second encoding unit 120 using this CELP encoding structure, the codebook from the shape index and gain circuit 126 of the codebook from the noise code book 121 is used. Gain index of is taken out. The shape index which is the UV data from the noise code field 121 is sent to the output terminal 107s via the switch 127s, and the gain index which is the UV data of the gain circuit 126 is output via the switch 127g. I am sent to (107g).

Here, the switches 127s and 127g and the switches 117 and 118 are controlled on / off by the V / UV determination result from the V / UV determination 115, and the switches 117 and 118 are controlled. Is on when the V / UV determination result of the voice signal of the frame to be transmitted is voiced sound (V), and switches 127s and 127g are on when the voice signal of the frame to be transmitted is silent (UV). Becomes

Next, with reference to FIG. 4, the more specific structure of the audio signal decoding apparatus shown in FIG. 2 is demonstrated. In FIG. 4, the same code | symbol is attached | subjected to the part corresponding to FIG.

In FIG. 4, the input terminal 202 is supplied with the vector quantized output of the LSP corresponding to the output from the output terminal 102 of FIGS. 1 and 3, and the index of the so-called codebook.

The index of this LSP is sent to the inverse vector quantizer 231 of the LSP of the LPC parameter reproducing unit 213, inverse vector quantized into LSP (line spectrum pair) data, and sent to the LSP interpolation circuits 232 and 233. After the LSP interpolation processing is performed, the LSP interpolation circuit 232 and the LSP to alpha conversion circuit 234 are for voiced sound (V), and the LSP interpolation circuit 233 and the LSP to alpha conversion circuit 235 are silent. It is for (UV). The LPC synthesis filter 214 separates the LPC synthesis filter 236 of the voiced sound portion and the LPC synthesis filter 237 of the unvoiced sound portion. In other words, the LPC coefficient interpolation is performed independently in the voiced and unvoiced portions, thereby preventing the adverse effects of interpolating LSPs having completely different properties in the transition portion from the voiced sound to the unvoiced sound or the transition portion from the unvoiced sound to the voiced sound. Doing.

In the input terminal 203 of FIG. 4, the weighted vector quantized code index data of the spectral envelope Am corresponding to the output from the terminal 103 on the encoder side of FIGS. 1 and 3 is supplied. Pitch data from the terminals 104 of FIGS. 1 and 3 are supplied to the input terminal 204, and V / UV determination data from the terminals 105 of FIGS. 1 and 3 are supplied from the input terminal 205. FIG. It is becoming.

The vector quantized index data of the spectral envelope Am from the input terminal 203 is sent to the inverse vector quantizer 212 to perform inverse vector quantization. As described above, the number of amplitude data of the inverse vector quantized envelope is a fixed number, for example, 44. The number of data is converted as if it were the harmonics of the number according to the pitch data. The number of data sent from the inverse vector quantizer 212 to the data number converter 270 may be converted into a number of data while remaining as a predetermined number.

In addition, pitch data at the input terminal 204 is supplied to the data converter 270 via the pitch converter 215, and the encoded pitch is supplied. Here, when pitch conversion is necessary, pitch conversion can be performed by the process described later in the pitch conversion unit 215. The amplitude data of the number corresponding to the set pitch of the spectral envelope of the LPC residual in the data converter 270 and the changed pitch data are sent to the sine wave synthesis circuit 215 of the voiced sound synthesis unit 211.

Here, in order to convert the number of amplitude data of the spectral envelope of the LPC residual in the data converter 270, various interpolation methods are considered, for example, the amplitude data of one block of the effective band on the frequency axis. On the other hand, dummy data for interpolating values from the last data in the block to the first data in the block are added to increase the number of data to N _F , or the left and right (first and last) data in the block are extended. O _S-fold of the band-limited data as dummy data by carrying out over sampling O _S times to obtain an amplitude data of the number of the O _S times the number ((m _MX +1 of) of (e.g. eight times) XO _S ) Amplitude data may be linearly interpolated, and further expanded to a large number of N _M pieces (for example, 2048 pieces), and the N _M data may be converted into data of the number M according to the pitch set by crushing.

In the data conversion section 270, since the shape of the spectral envelope is not changed, only the upright position of the harmonics can be changed. For this reason, phonology is immutable.

Here, as an example of the operation in the data converter 270, the case where the frequency F ₀ = f _s / L at the time of the pitch lag L is converted into Fx will be described. f _s is a sampling frequency, for example, f _s = 8 kHz = 8000 Hz.

At this time, the pitch frequency F ₀ = 8000 / L, and n = L / 2 is set up to 4000 Hz. In the 3400 Hz width of a typical audio band, it is about (L / 2) x (3400/4000). This is converted into a predetermined number, for example, 44 by the data number conversion or dimension conversion described above, and then vector quantization is performed.

If the interframe difference is taken before the vector quantization of the spectrum at the time of encoding, the interframe difference is decoded after the inverse vector quantization here, so that data number conversion is performed to obtain data of the spectral envelope.

The sine wave synthesis circuit 215 is supplied with the V / UV determination data at the input terminal 205 in addition to the spectral envelope amplitude data and the pitch data of the LPC residual in the data converter 270. In the sine wave combining circuit 215, the LPC residual data is taken out and sent to the adder 218.

In addition, the envelope data from the inverse vector quantizer 212, the pitch from the input terminal 204, and the V / UV determination data from the input terminal 205 are synthesized to add noise to the voiced sound (V) portion. It is sent to the circuit 216. The output from the noise synthesis circuit 216 is sent to the adder 218 via the weighted overlap addition circuit 217. This creates excitation (excitation), which is the input of voiced sound to the LPC synthesis filter by sine wave synthesis. Considering the fact that the sound quality may change suddenly and unnaturally in the case of UV and unvoiced sound, and the LPC synthesis filter input, ie, excitation, of the voice part is based on parameters such as pitch and spectrum spectrum. Noise in consideration of the rope amplitude, the maximum amplitude in the frame, and the residual signal level is applied to the voiced sound portion of the LPC residual signal.

The addition output from the adder 218 is sent to the synthesis filter 236 for the voiced sound of the LPC synthesis filter 214, and is subjected to LPC synthesis processing, thereby making time waveform data, and the post filter 238v for the voiced sound. After filtering, it is sent to the adder 239.

Next, the shape index and the gain index as the UV data from the output terminals 107s and 107g of FIG. 3 are taken out to the input terminals 207s and 207g of FIG. 3, respectively, and are sent to the unvoiced sound synthesis unit 220. FIG. The shape index from the terminal 207s is sent to the noise coding field 221 of the unvoiced synthesizer 220, and the gain index from the terminal 207g is sent to the gain circuit 222, respectively. The representative value output read out from the noise code field 221 is a noise signal component corresponding to the excitation vector, i.e., the LPC residual of the unvoiced sound, which becomes the amplitude of the predetermined gain in the gain circuit 222, and the windowing circuit 223 Is sent to a windowing process for smoothing the connection with the voiced sound portion.

The output from the windowing circuit 223 is sent to the synthesis filter 237 for UV (silent sound) of the LPC synthesis filter 214 as the output from the unvoiced sound synthesis unit 220. In the synthesis filter 237, the LPC synthesis processing is performed to form time waveform data of the unvoiced sound portion, which is filtered by the unvoiced post filter 238u, and then sent to the adder 239. FIG.

In the adder 239, the time waveform signal of the voiced sound portion from the voiced sound post filter 238v and the time waveform data of the unvoiced sound portion from the unvoiced post filter 230u are added and taken out from the output terminal 201.

Pitch conversion processing performed in the pitch conversion unit 119 included in the audio encoding apparatus described with reference to FIGS. 1 and 3, and pitch conversion unit included in the audio decoding apparatus described with reference to FIGS. The pitch conversion processing performed at 240 will be described. This example is configured to enable pitch conversion of speech both during encoding and during decoding. When pitch conversion is to be performed at the time of encoding, the corresponding processing is performed by the pitch conversion unit 119 included in the speech encoding apparatus. When pitch conversion is to be performed at the time of decoding, the corresponding processing is performed by the pitch conversion unit 240 included in the speech decoding apparatus. Therefore, basically, if there is a pitch conversion section on either side of the audio encoding device or the audio decoding device, the pitch conversion processing described in this example can be executed. The speech signal pitch-transformed at the time of encoding in the speech encoding apparatus may be further pitch-converted at the time of decoding in the speech decoding apparatus.

Hereinafter, the detail of the process in a pitch conversion part is demonstrated. The pitch conversion processing performed by the pitch conversion unit 119 included in the audio encoding device and the pitch conversion processing performed by the pitch conversion unit 215 in the audio decoding device are basically the same. In each of the converters 119 and 240, the supplied pitch data is converted. In this example, the pitch data supplied to each pitch converter 119 is a pitch rack (period) as described with reference to FIGS. 1 to 4. The pitch rack is converted into other data by arithmetic processing, and pitch conversion is performed.

The specific processing of this pitch conversion can be selected from the first processing to the ninth processing described below, i.e., in nine processing states. One of these processing states is set based on control performed by a controller or the like included in the encoding apparatus or the decoding apparatus. The pitch represented by the formula in the description of the following processing represents the period. In the arithmetic processing in the actual conversion section, the corresponding processing is performed with data of the number of harmonics.

First treatment

This process is a process of multiplying the input pitch by an integer. The input pitch pch_in is multiplied by an integer K ₁ to calculate the output pitch pch_out. The calculation is shown in the following equation.

[Equation 1]

pch_out = K ₁ pch_in

Here, if the value of the constant K ₁ is set to satisfy the relationship of 0 <K ₁ <1, the frequency becomes high and can be changed to a high sound. When the value of the constant K ₁ is set as a value so as to satisfy the relationship of K ₁ > 1, the frequency becomes low and can be changed to a low sound.

Second treatment

This process is a process of making the output pitch constant regardless of the input pitch. Always set the appropriate constant P ₂ as the output pitch (pch_out). The calculation is shown in the following expression.

[Equation 2]

pch_out = P ₂

By setting it to a constant pitch in this way, it can be converted into monotonous artificial sound.

Third treatment

This process makes the output pitch pch_out equal to the sum of a sine wave having a suitable constant P ₃ set in advance and an appropriate amplitude A ₃ frequency F ₃ . The calculation is shown in the following equation.

[Equation 3]

pch_out = P ₃ + A ₃ sin (2πF ₃ t _(n) )

In Equation 3, n is the number of frames, and t _(n) is set by the following equation as a discrete time in the frame.

[Equation 4]

t _(n) = t _(n-1) + Δt

By adding a square wave to the fixed constant pitch as described above, vibration may be applied to the artificial sound.

Fourth treatment

This process is to make the output pitch pch_out by adding the input pitch pch_in and the same random number [-A ₄ , A ₄ ]. The calculation is shown in the following equation.

[Equation 5]

pch_out = pch_in + r _(n)

Here, r _(n) is a random number set for each frame n. For each processing frame, the same random number [-A ₄ , A ₄ ] is generated, and an addition process is performed. By doing this, you can convert the sound into a rattling sound.

Fifth treatment

This process is a process of making the output pitch pch_out by adding the sinusoids of the appropriate amplitude A ₅ and the frequency F ₅ to the input pitch pch_in. The calculation is shown in the following equation.

[Equation 6]

pch_out = pch_in + A ₅ sin (2πF ₅ t _(n) )

In this equation (6 ₎ , n is the number of frames, t _(n) is the discrete time in the frame, and is set by the above equation (4). In this way, vibration is applied to the input voice. In this case, by changing the frequency F ₅ to a small value (that is, lengthening the period), the conversion to the sound with the up and down is performed.

6th treatment

This processing subtracts the input pitch pch_in from the appropriate constant P _{6 as the} output pitch pch_out. The calculation is shown in the following expression.

[Equation 7]

pch_out = P ₆ -pch_in

In this way, the change in pitch is inverse to the input voice. For example, a change in the ending is converted into a voice that is inverted.

7th treatment

This process outputs avg_pch obtained by smoothing (averaging) the input pitch pch_in with an appropriate time constant τ ₇ (where time constant τ ₇ is in the range of 0 <τ ₇ <1). ) The calculation is shown in the following expression (8).

[Equation 8]

avg＿pch = (1-τ ₇ ) avg＿pch + τ ₇ pch＿in

pch＿out = avg＿pch

Here, for example, by setting 0.05 as tau ₇ , the average value of the past 20 frames becomes avg_pch, and the value becomes the output pitch. In this way, the sound is converted into a sulky persimmon without rising or falling.

8th treatment

This process is performed by avg_pch obtained by smoothing (averaging) the input pitch pch_in with an appropriate time constant τ ₈ (where the time constant τ ₈ is in the range of 0 <τ ₈ <1). pch_in). The difference is multiplied by a suitable factor K ₈ , where K ₈ is an integer. The product of this result is added to the input pitch pch_in as the emphasis component, and is set as the output pitch pch_out. The calculation is shown in the following expression (9).

[Equation 9]

avg＿pch = (1-τ ₈ ) avg＿pch + τ ₈ pch＿in

pch＿out = pch＿in + K ₈ (pch＿in-avg＿pch)

In this way, the pitch conversion is performed with the emphasis component added to the input voice. In effect, the conversion to the modulated sound is performed.

9th treatment

This is a mapping process of converting the input pitch pch_in into the closest fixed pitch data contained in the pitch table prepared in advance in the pitch converting section. In this case, for example, as fixed pitch data in the pitch table, data of frequency intervals corresponding to the scale are prepared and converted into data of the scale closest to the input pitch pch_in.

By performing any of the pitch conversion processing from the first to the ninth processing described above in the pitch conversion unit 119 in the encoding apparatus or the pitch conversion unit 240 in the decoding apparatus, the number of harmonics during decoding is controlled. Only pitch data is converted. Therefore, it is possible to simply change the pitch without changing the phonation of the voice.

An example in which the voice encoding device and the voice decoding device described above are applied to a telephone apparatus will be described with reference to FIGS. 5 and 6. First, FIG. 5 shows an example in which the voice encoding device is applied to a transmission system of a radiotelephone device (such as a cellular phone set). The voice signal collected by the microphone 301 is amplified by the amplifier 302, converted into a digital signal by the analog / digital converter 303, and sent to the voice encoder 304. This audio encoding unit 304 corresponds to the audio encoding apparatus described with reference to FIGS. 1 and 3. If necessary, the pitch conversion processing is performed in the pitch conversion unit (corresponding to the pitch conversion unit 119 in Figs. 1 and 3) included in the coding unit 304. Each data encoded by the speech encoder 304 is sent to the transmission path encoder 305 as an output signal of the encoder 304. In the transmission path encoding unit 305, a so-called channel encoding process is performed. The output signal is sent to the modulation circuit 306 to be modulated, sent to the antenna 309 via the digital-to-analog converter 307 and the high frequency amplifier 308, and wirelessly transmitted.

6 shows an example in which the voice decoding device is applied to the transmission system of the radiotelephone device. The signal received at the antenna 311 is amplified by the high frequency amplifier 312 and sent to the demodulation circuit 314 via the analog / digital converter 313. The demodulated signal is sent to the transmission path decoding unit 315. In this transmission path decoding unit 315, the voice is subjected to channel decoding processing, and the transmitted voice signal is extracted. The extracted voice signal is sent to the voice decoder 316. This voice decoding unit 316 corresponds to the voice decoding apparatus described with reference to FIGS. 2 and 4. If necessary, pitch conversion processing is performed in the pitch conversion unit (corresponding to the pitch conversion unit 215 in Figs. 2 and 4) in the coding unit 316. The audio signal decoded by the voice decoding unit 316 is sent to the digital / analog converter 317 as an output signal of the decoding unit 316, and analog voice processing is performed by the amplifier 318, and then to the speaker 319. Sent and radiated as voice.

Of course, the present invention can be applied to devices other than such a radiotelephone apparatus. That is, the present invention can be applied to various devices incorporating the audio encoding device described in FIG. 1 and the like and handling various audio signals incorporating the audio decoding device described in FIG.

Incidentally, a process corresponding to the processing in the pitch converter 119 of this example on a recording medium (optical disk, magneto-optical disk, magnetic tape, etc.) in which the processing program for executing the audio encoding process described in Figs. In the case where a program is recorded and a processing program read from this medium is executed and encoded in a computer apparatus or the like, the same pitch conversion processing can be executed. Similarly, a processing program corresponding to the processing in the pitch converting unit 240 of this example is recorded on a recording medium on which a processing program for executing the audio decoding process described with reference to Figs. 2 and 4 is recorded and read from this medium. In the case where the processing program is executed in the computer apparatus or the like to decode, the same pitch conversion processing can be executed.

According to the speech encoding method of the present invention, the pitch component of the sinusoidal encoded speech encoded data is changed to a predetermined calculation process and the pitch is converted. As a result, it is possible to convert and decode only the pitch accurately without changing the phonation of the input voice by simple arithmetic processing.

In this case, data number conversion is performed in which the number of harmonics is a predetermined number. As a result, pitch conversion based on the decoded data is simply performed.

In the case of performing this data number conversion, the data number conversion processing is performed by interpolation processing using the oversampling operation. As a result, data number conversion is performed by a simple process using the oversampling operation.

When pitch conversion is performed at the time of encoding, a predetermined coefficient is multiplied by the pitch component of the sinusoidal coded speech coded data to perform the pitch conversion. As a result, for example, a pitch conversion process for changing the timbre of the input voice becomes possible.

In addition, when pitch conversion is performed at the time of encoding, the pitch component of the sinusoidal encoded audio coded data is converted into a fixed value, and is always converted to a constant pitch. For example, therefore, the pitch of the input speech can be converted to monotonous artificial speech.

In the case of converting to a constant pitch, the sinusoidal data of a predetermined frequency is added to the constant pitch-converted data. As a result, for example, the sound can be converted into a voice having vibration up and down about a constant pitch.

In addition, when pitch conversion is performed at the time of encoding, the pitch component of the sine wave analysis encoded speech encoded data is subtracted from a predetermined integer value, and pitch conversion is performed. As a result, for example, it is possible to convert the pitch into a pitch such that an effect such as inversion of the mother of the input voice is reversed.

In addition, when pitch conversion is performed at the time of encoding, a predetermined random number is added to the pitch component of the sinusoidal encoded audio coded data, and the pitch transform is performed. As a result, it is possible to convert the speech into pitches that seem to change irregularly.

When pitch conversion is performed at the time of encoding, sinusoidal data of a predetermined frequency is added to the pitch component of speech encoded data encoded using sine wave analysis encoding, and pitch conversion is performed. As a result, for example, it becomes possible to convert to the speech obtained by applying vibration to the input speech.

In addition, when pitch conversion is performed at the time of encoding, the average value of the pitch components of the sinusoidal coded speech coded data is calculated, and this average value is used as the speech coded data and pitch transformed. As a result, for example, it becomes possible to convert the input voice into a voice with reduced rise and fall.

When pitch conversion is performed at the time of encoding, the average value of the pitch components of the sinusoidal coded speech coded data is calculated, and the difference between the speech coded data and the average value is added to the speech coded data to perform the pitch conversion. As a result, for example, it is possible to convert a voice which is emphasized by the rising and falling of the input voice and is effectively modulated.

When pitch conversion is performed at the time of encoding, the pitch component of the sinusoidal encoded audio coded data is converted into data of a pitch conversion table prepared in advance, and converted to the pitch of the step set in the pitch conversion table. As a result, for example, the conversion such that the pitch of the input voice is normalized to the pitch of a certain scale can be performed.

According to the speech encoding method of the present invention, the sine wave analysis encoded pitch component is changed to a predetermined calculation process. As a result, only the pitch of the decoded speech can be accurately converted without changing the phonation of the speech by simple arithmetic processing.

In this case, after the pitch component is changed, data number conversion is performed from the predetermined number to the number of harmonics. As a result, decoding by the changed pitch component is simply performed.

In the case of performing this data number conversion, the data number conversion processing is performed by interpolation processing using the oversampling operation. In a simple process using the oversampling operation, data number conversion is performed.

When pitch conversion is performed at the time of decoding, a pitch coefficient is multiplied by a predetermined coefficient multiplied by the pitch component of the sinusoidal coded speech coded data. As a result, pitch conversion processing such as changing the tone of the decoded voice can be performed.

When pitch conversion is performed at the time of decoding, the pitch component of the sinusoidal encoded audio coded data is converted to a fixed value, and is always converted to a constant pitch. For example, therefore, the pitch of the decoded voice can be converted into a monotonous artificial voice.

In the case of conversion to a constant pitch, data of a sinusoidal wave of a predetermined frequency is added to the data converted to a constant pitch. As a result, it becomes possible to convert, for example, to a voice having a vibration up and down around a constant pitch.

In addition, when pitch conversion is performed at the time of decoding, the pitch component of the sine wave analysis encoded speech encoded data is subtracted from a predetermined integer value, and pitch conversion is performed. As a result, for example, it is possible to convert the pitch into an effect such that the mother's intonation of the decoded voice is reversed.

When pitch conversion is performed at the time of decoding, a predetermined random number is added to the pitch component of the sinusoidal encoded audio coded data, and pitch conversion is performed. As a result, for example, it becomes possible to convert the pitch into which an internation of a decoded voice, etc. changes irregularly.

In the case where pitch conversion is performed at the time of decoding, sinusoidal data of a predetermined frequency is added to the pitch component of speech encoded data by sinusoidal analysis encoding, and pitch conversion is performed. As a result, for example, it becomes possible to convert the voice to be obtained by applying vibration to the decoded voice.

When pitch conversion is performed at the time of decoding, the average value of the pitch components of the sinusoidal coded speech coded data is calculated, and the average value is used as the pitch coded speech coded data. As a result, for example, it becomes possible to convert the intonation of the decoded speech into the reduced speech.

When pitch conversion is performed at the time of decoding, the average value of the pitch components of the sinusoidal coded speech coded data is calculated, and the difference between the speech coded data and the average value is added to the speech coded data to be pitch transformed. As a result, for example, the intonation of the decoded voice is emphasized and it is possible to convert the voice into a modulated effect.

In the case where pitch conversion is performed at the time of decoding, the pitch component of the sinusoidal encoded coded speech coded data is converted into data of a pitch conversion table prepared in advance, and converted to the pitch of the steps set in this pitch conversion table. As a result, for example, a conversion such as normalizing the pitch of the decoded voice to a pitch of a certain scale can be performed.

According to the speech coding apparatus of the present invention, there is provided a pitch converting means for changing a sine wave analysis encoded pitch component into a predetermined calculation process. The pitch component of the sinusoidal encoded data can be converted and decoded accurately by only the pitch without changing the phonation of the input voice with a simple processing configuration by the conversion process.

In this case, data number conversion is performed in which the number of harmonics is a predetermined number. As a result, encoding can be performed with a simple processing structure. Further, pitch conversion based on the encoded data is simply performed.

This data number conversion processing is performed by interpolation processing by a band limited oversampling filter. As a result, data number conversion processing is performed with a simple processing structure using an oversampling filter.

According to the speech decoding apparatus of the present invention, the pitch component of the sinusoidal encoded data is converted into a pitch converting means, and the decoding process is performed by the speech decoding means by the encoded sinusoidal encoded data and the encoded data based on the linear prediction residual. All. Therefore, it is possible to accurately convert only the pitch of the decoded voice without changing the phonation of the voice with a simple processing configuration.

In this case, data number conversion is performed from the predetermined number of harmonics. As a result, the converted pitch is decoded by a simple processing structure as much as the number of harmonics is changed.

This data number conversion processing is performed by interpolation processing using a band limited oversampling filter. As a result, with the simple processing structure using the oversampling filter, data number conversion during decoding is performed.

The telephone apparatus according to the present invention also has a pitch converting means for converting pitch components of the data encoded by the sine wave analysis encoding means. Therefore, the pitch component of the audio data to be transmitted with a simple configuration can be converted into a desired state.

According to the pitch conversion method of the present invention, a sine wave analysis of a speech signal is performed to multiply a predetermined coefficient by the encoded pitch component data, and pitch conversion is performed. As a result, for example, conversion such as changing the timbre of the input voice is performed simply.

In addition, according to the pitch conversion method of the present invention, a sine wave analysis of a speech signal is performed to convert the encoded pitch component data into a fixed value, which is always converted to a constant pitch. As a result, for example, the pitch of the input voice is converted into a monotonous artificial voice.

Further, according to the pitch conversion method of the present invention, the speech encoded data encoded by the sine wave analysis encoding is subtracted from the predetermined integer value, and the pitch conversion is performed. As a result, for example, it becomes possible to convert the pitch into a pitch such that an effect such as the intonation of the input voice is reversed.

According to the medium of the present invention, a processing program for converting pitch components of speech encoded data encoded by sine wave analysis encoding is recorded on a medium on which an encoded program is recorded. By executing this processing program, it is possible to accurately convert and encode only pitches without changing the phonation of the input voice.

Further, according to the medium of the present invention, a pitch conversion processing program for converting pitch components of sine wave analysis decoded data is recorded on a medium on which a decoding program is recorded. Therefore, by executing this processing program, it is possible to accurately convert only the pitch of the decoded speech without changing the phonation of the input speech.

Although embodiments of the present invention have been described with reference to the drawings, the present invention is not limited to the above-described embodiments, and various changes and modifications are defined in the claims by the person skilled in the art. Possible without departing from the scope.

1 is a block diagram showing a basic configuration of an example of an audio encoding apparatus according to an embodiment of the present invention.

2 is a block diagram showing the basic configuration of an audio signal decoding apparatus according to an embodiment of the present invention.

3 is a block diagram illustrating a more detailed configuration of the audio signal encoding apparatus of FIG. 1.

4 is a block diagram showing a more detailed configuration of the audio signal decoding apparatus of FIG.

5 is a block diagram showing an example applied to a transmission system of a radiotelephone apparatus.

6 is a block diagram showing an example applied to the reception system of the radiotelephone apparatus.

* Explanation of symbols on the main parts of the drawings

110. First Coder 111. LPC Inverse Filter

113. LPC analysis and quantization unit 114. Sine wave analysis encoding unit

115.V / UV Determination 119. Pitch Conversion Unit

120. Second Coder 121. Noise Code

122. Weighted synthetic filter 123. Subtractor

124. Distance calculation circuit 125. Acoustic weight filter

211. Voiced speech synthesis circuits 212. Inverse vector quantizers

215. Pitch converter 270. Data converter

Claims (36)

Dividing the speech signal into predetermined coding units on the time axis; Obtaining a linear prediction residual for each split coding unit; and In the speech encoding method comprising the step of performing a sinusoidal analysis encoding based on the linear prediction residual, And changing a pitch component of said sine wave analysis encoded speech encoding data to a predetermined calculation process. The method of claim 1, The encoding process is performed by harmonic encoding, and data number conversion is performed in which the number of harmonics is a predetermined number. The method of claim 2, And the data number conversion processing is performed by an interpolation processing using an oversampling operation. The method of claim 1, A speech encoding method characterized by performing a pitch conversion by multiplying a pitch component of a sine wave analysis encoded speech encoding data by a predetermined coefficient. The method of claim 1, A speech encoding method comprising converting a pitch component of a sinusoidal encoded speech encoded data into a fixed value and always converting the pitch component into a fixed pitch. The method of claim 5, And a sine wave of a predetermined frequency is added to the data of the predetermined pitch. The method of claim 1, And a pitch conversion is performed by subtracting the pitch component of the sine wave analysis encoded speech encoded data from a predetermined integer value. The method of claim 1, And a predetermined random number is added to a pitch component of said sine wave analysis encoded speech encoding data, and pitch conversion is performed. The method of claim 1, And the sine wave data of a predetermined frequency is added to the pitch component of the speech encoded data by the sine wave analysis encoding, and the pitch encoding is performed. The method of claim 1, And an average value of pitch components of speech encoded data is calculated by the sine wave analysis encoding, and the average is converted into pitch encoded speech encoded data. The method of claim 1, And an average value of pitch components of the speech encoded data encoded by the sine wave analysis encoding, and a difference between the speech encoded data and the average value is added to the speech encoded data to perform pitch conversion. The method of claim 1, And a pitch component of the sine wave analysis encoded speech encoding data is converted into data of a pitch conversion table prepared in advance, and converted into a pitch of a step set in the pitch conversion table. A speech decoding method in which a speech signal is decoded on the basis of linear predictive residual data of a predetermined coding unit on a time axis and sine wave analysis encoded data, And changing a pitch component of the sine wave analysis encoded data by a predetermined calculation process. The method of claim 13, And converting the number of harmonics to a predetermined number in a coding process using harmonics coding after the pitch component is changed to a predetermined arithmetic processing. The method of claim 14, And said data number conversion processing is performed by interpolation processing using oversampling operation. The method of claim 13, A speech decoding method comprising performing a pitch conversion by multiplying a pitch component of a sinusoidal encoded speech encoded data by a predetermined coefficient. The method of claim 13, A speech decoding method comprising converting a pitch component of a sinusoidal coded speech coded data into a fixed value and always converting the pitch component into a fixed pitch. The method of claim 17, And a sine wave of a predetermined frequency is added to the data of the predetermined pitch. The method of claim 13, And a pitch conversion is performed by subtracting the pitch component of the sine wave analysis encoded speech encoded data from a predetermined integer value. The method of claim 13, And a predetermined random number is added to a pitch component of said sine wave analysis encoded speech encoding data, and pitch conversion is performed. The method of claim 13, And the sine wave data of a predetermined frequency is added to the pitch component of the speech encoded data by the sine wave analysis encoding, and the pitch decoding is performed. The method of claim 13, And an average value of pitch components of speech encoded data is calculated by the sine wave analysis encoding, and the average is converted into pitch encoded speech encoded data. The method of claim 13, And an average value of pitch components of the speech encoded data encoded by the sine wave analysis encoding, and a difference between the speech encoded data and the average value is added to the speech encoded data to perform pitch conversion. The method of claim 13, And a pitch component of the sine wave analysis encoded speech encoding data is converted into data of a pitch conversion table prepared in advance, and converted into a pitch of a step set in the pitch conversion table. Linear predictive residual detection means for obtaining a linear predictive residual from an input speech signal in a predetermined coding unit on a time axis; Sine wave analysis encoding means for sine wave analysis encoding the linear predicted residual detected by the linear predictive residual detection means; And a pitch conversion means for converting a pitch component of the data encoded by the sine wave analysis encoding means. The method of claim 25, And a data number conversion for setting the number of harmonics used for harmonic encoding to a predetermined number is performed by said sinusoidal encoding encoding means. The method of claim 26, And said data number conversion processing is performed by interpolation processing using a band limited oversampling filter. A speech decoding apparatus for decoding a speech signal on the basis of linear predictive residual data of a predetermined coding unit on a time axis and sine wave analysis encoded data, Pitch conversion means for converting a pitch component of said sine wave analysis encoded data; And a speech decoding means for performing a decoding process on the sine wave analysis coded data and the linear predictive residual data converted by the pitch converting means. The method of claim 28, And a data number conversion for setting a predetermined number of harmonics used for harmonic encoding on the basis of the converted pitch component data. The method of claim 29, And the data number conversion processing is performed by interpolation processing using a band-limited oversampling filter. Linear predictive residual detection means for obtaining a linear predictive residual from an input speech signal in a predetermined coding unit on a time axis; Sinusoidal stone encoding means for encoding sinusoidal wave analysis encoding the linear predictive residual detected by the linear predictive residual detection means; Pitch conversion means for converting pitch components of the data encoded by the sine wave analysis encoding means; And transmission means for transmitting the analysis coded data and the linear prediction residual data converted by the pitch conversion means to a predetermined transmission line. A pitch conversion method comprising sine wave analysis of a speech signal and multiplying a predetermined coefficient by the encoded pitch component data to perform a pitch conversion. Sine wave analysis of the speech signal to convert the data of the encoded pitch component to a fixed value, the pitch conversion method comprising the step of always converting to a constant pitch A pitch conversion method comprising subtracting data of a pitch component encoded by sine wave analysis encoding from a predetermined integer value and performing pitch conversion. Processing for classifying the input audio signal into predetermined coding units on the time axis; A process of detecting a linear prediction residual in each of the divided coding units; A medium on which a program for performing a process for encoding sinusoidal analysis of the linear prediction residuals is recorded. And a processing program for converting a pitch component of said sine wave analysis encoded speech encoded data. A medium in which a processing program for decoding a speech signal is recorded on the basis of linear predictive residual data and sine wave analysis encoded data of a predetermined coding unit on a time axis, A medium for recording a pitch conversion processing program for converting pitch components of said sine wave analysis encoded data;

Images