Classifications

• • G—PHYSICS
  • G10—MUSICAL INSTRUMENTS; ACOUSTICS
  • G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
  • G10L19/00—Speech or audio signals analysis-synthesis techniques for redundancy reduction, e.g. in vocoders; Coding or decoding of speech or audio signals, using source filter models or psychoacoustic analysis
  • G10L19/02—Speech or audio signals analysis-synthesis techniques for redundancy reduction, e.g. in vocoders; Coding or decoding of speech or audio signals, using source filter models or psychoacoustic analysis using spectral analysis, e.g. transform vocoders or subband vocoders
• • G—PHYSICS
  • G10—MUSICAL INSTRUMENTS; ACOUSTICS
  • G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
  • G10L19/00—Speech or audio signals analysis-synthesis techniques for redundancy reduction, e.g. in vocoders; Coding or decoding of speech or audio signals, using source filter models or psychoacoustic analysis
  • G10L19/02—Speech or audio signals analysis-synthesis techniques for redundancy reduction, e.g. in vocoders; Coding or decoding of speech or audio signals, using source filter models or psychoacoustic analysis using spectral analysis, e.g. transform vocoders or subband vocoders
  • G10L19/032—Quantisation or dequantisation of spectral components
• • G—PHYSICS
  • G10—MUSICAL INSTRUMENTS; ACOUSTICS
  • G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
  • G10L19/00—Speech or audio signals analysis-synthesis techniques for redundancy reduction, e.g. in vocoders; Coding or decoding of speech or audio signals, using source filter models or psychoacoustic analysis
  • G10L19/04—Speech or audio signals analysis-synthesis techniques for redundancy reduction, e.g. in vocoders; Coding or decoding of speech or audio signals, using source filter models or psychoacoustic analysis using predictive techniques
  • G10L19/08—Determination or coding of the excitation function; Determination or coding of the long-term prediction parameters
  • G10L19/087—Determination or coding of the excitation function; Determination or coding of the long-term prediction parameters using mixed excitation models, e.g. MELP, MBE, split band LPC or HVXC
• • G—PHYSICS
  • G10—MUSICAL INSTRUMENTS; ACOUSTICS
  • G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
  • G10L19/00—Speech or audio signals analysis-synthesis techniques for redundancy reduction, e.g. in vocoders; Coding or decoding of speech or audio signals, using source filter models or psychoacoustic analysis
  • G10L19/04—Speech or audio signals analysis-synthesis techniques for redundancy reduction, e.g. in vocoders; Coding or decoding of speech or audio signals, using source filter models or psychoacoustic analysis using predictive techniques
  • G10L19/16—Vocoder architecture
  • G10L19/18—Vocoders using multiple modes
  • G10L19/20—Vocoders using multiple modes using sound class specific coding, hybrid encoders or object based coding
• • G—PHYSICS
  • G11—INFORMATION STORAGE
  • G11B—INFORMATION STORAGE BASED ON RELATIVE MOVEMENT BETWEEN RECORD CARRIER AND TRANSDUCER
  • G11B20/00—Signal processing not specific to the method of recording or reproducing; Circuits therefor
  • G11B20/10—Digital recording or reproducing
• • G—PHYSICS
  • G10—MUSICAL INSTRUMENTS; ACOUSTICS
  • G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
  • G10L25/00—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00
  • G10L25/93—Discriminating between voiced and unvoiced parts of speech signals

Definitions

• the present invention encodes an audio signal and records it on a transmission or recording medium, and receives or decodes the audio signal on a decoding side.
• the present invention relates to a recording medium on which a recorded code string is recorded.
• Sub-band coding (Sub-Band Coding: SBC), which is a non-blocking frequency band division method that transforms signals, and converts signals on the time axis to signals on the frequency axis (spectral conversion) to convert to multiple frequency bands Blocked frequency band division, in which division is performed and encoding is performed for each band, so-called conversion encoding can be used.
• SBC Sub-band Coding
• a high-efficiency coding method combining the above-described band division coding and transform coding is also considered. In this case, for example, after performing band division by the above band division coding, The signal for each band is spectrally transformed into a signal on the frequency axis, and encoding is performed for each of the spectrally transformed bands.
• the input acoustic time-series signal is blocked by a frame of a predetermined unit time, and discrete Fourier transform (DFT), discrete
• DFT discrete Fourier transform
• DCT discrete Fourier transform
• MDCT discrete Fourier transform
• the filter or the spectrum transform By quantizing the signal divided for each band by the filter or the spectrum transform in this way, the band in which the quantization noise occurs can be controlled, and the characteristics t such as the masking effect can be used. More efficient encoding can be performed audibly. Further, if the normalization is performed for each band before quantization, for example, with the maximum value of the absolute value of the signal component in that band, more efficient coding can be performed. .
• band division is performed in consideration of human auditory characteristics.
• the audio signal is divided into a plurality of bands, such as 32 bands, with a bandwidth generally called a critical band, in which the higher the frequency band, the wider the bandwidth.
• a predetermined bit allocation or bit allocation and bit allocation
• an appropriate bit allocation for each band are performed. Is performed. For example, when encoding the coefficient data obtained by the above MDCT processing, the coding is performed using the number of bits that are adaptively assigned to the MDCT coefficient data for each band obtained by the MDCT processing for each block. Will be performed.
• a spectrum with high energy that is, a tone component T is locally separated from the spectrum on the frequency axis as shown in FIG. 1A.
• the noise component excluding the tone component has a spectrum as shown in Fig. 1B. Then, each of the tone component and the noise component is quantized with sufficient and appropriate accuracy.
• the present invention has been proposed in view of the above situation, and has an audio signal encoding method and an apparatus for suppressing an encoding efficiency from being deteriorated by a tone component existing at a local frequency.
• An object of the present invention is to provide a signal decoding method and an apparatus therefor, and an audio signal encoding program, an audio signal decoding program, or a recording medium on which a code sequence encoded by the audio signal encoding device is recorded. It is assumed that.
• An audio signal encoding method is the audio signal encoding method for encoding an audio time series signal, wherein a tone component encoding step of extracting and encoding a tone component signal from the audio time series signal,
• the tone component encoding step includes a residual component encoding step of encoding a residual time series signal obtained by extracting the tone component signal from the acoustic time series signal.
• a tone component signal is extracted from an audio time series signal, and the tone component signal and a residual time series signal obtained by extracting a tone component signal from the audio time series signal are encoded.
• the sound signal decoding method extracts a tone component signal from an acoustic time-series signal, encodes the tone component signal, and further comprises:
• An audio signal decoding method for inputting a code sequence obtained by encoding a residual signal obtained by extracting a two-tone component signal and decoding the code sequence, wherein the code sequence decomposes the code sequence
• a tone component signal is extracted from an acoustic time-series signal, and the tone component signal and a residual time-series signal obtained by extracting a tone component signal from the acoustic time-series signal are encoded. And decode the audio sequence to recover the acoustic time-series signal.
• the audio signal encoding method in the audio signal encoding method for encoding an audio time series signal, comprises: a frequency band division step of dividing the audio time series signal into a plurality of frequency bands; A tone component encoding step of extracting and encoding a tone component signal from the acoustic time series signal of one frequency band; and the acoustic time series signal of at least one frequency band in the tone component encoding step. And a residual component encoding step of encoding a residual time-series signal obtained by extracting the tone component signal from the above.
• a tone component signal is extracted from an audio time series signal for at least one frequency band of an audio time series signal divided into a plurality of frequency bands, and the tone component signal is extracted. And a residual time-series signal obtained by extracting a tone component signal from the acoustic time-series signal.
• the acoustic time-series signal is divided into a plurality of frequency bands, and the audio time-series signal is divided into at least one frequency band.
• a tone sequence signal is extracted and encoded, and a code sequence is input in which a residual time-series signal obtained by extracting the tone component signal from the acoustic time-series signal of at least one frequency band is encoded.
• An audio signal decoding method for decoding the code sequence wherein the code sequence decomposition process for decomposing the code sequence and the code sequence decomposition process for at least one frequency band are performed.
• Additive synthesis with signal It has a summing step of obtaining a decoding signal, a band synthesizing step for restoring the acoustic time-series signal by band synthesizing decodes signals for each band Te.
• a tone component signal is extracted from an audio time-series signal for at least one frequency band of an audio time-series signal divided into a plurality of frequency bands, and the tone component signal is extracted.
• the decoding unit decodes a code sequence formed by encoding the residual time-series signal obtained by extracting the tone component signal from the audio time-series signal, and restores the acoustic time-series signal.
• the audio signal encoding method is the audio signal encoding method for encoding an audio time-series signal, wherein a tone component signal is extracted from the audio time-series signal, and the tone component signal is encoded.
• the above second sound No. compares the coding efficiency of the coding process, and a coding efficiency determining step of selecting a good code sequence coding efficiency.
• a tone component signal is extracted from an audio time-series signal, and a residual obtained by extracting a tone component signal from the tone component signal and the audio time-series signal.
• a code sequence with good coding efficiency is selected from the code sequence with the second audio signal coding step of coding the sequence signal.
• the sound signal decoding method includes: extracting a tone component signal from an audio time-series signal; encoding the tone component signal; and generating the tone component signal from the audio time-series signal.
• the toe obtained in the code string decomposition step Component decoding process for generating a tone component time-series signal according to the residual component information, and residual component decoding for generating a residual component time-series signal according to the residual component information obtained in the code decomposition process. Recovering the acoustic time-series signal by a first acoustic signal decoding step having a step of adding and combining the tone component time-series signal and the residual component time-series signal.
• the audio time-series signal is restored by a second audio signal decoding step corresponding to the second audio signal encoding step.
• a tone component signal is extracted from an audio time series signal, and a residual time series signal obtained by extracting the tone component signal and the tone component signal from the audio time series signal.
• a first audio signal encoding step of encoding the audio time-series signal by a first encoding method of encoding the audio time-series signal by a first encoding method of encoding the audio time-series signal by a second encoding method The selected code stream having a high coding efficiency is input from the code stream of the second audio signal coding step to be decoded, and the corresponding decoding is performed on the coding side.
• An audio signal encoding device is an audio signal encoding device that encodes an audio time-series signal, comprising extracting and encoding a tone component signal from the time-series signal. And a residual component encoding unit that encodes a residual time-series signal in which the tone component signal is extracted from the acoustic time-series signal by the tone component encoding unit. It is characterized by.
• Such an audio signal encoding apparatus extracts a tone component signal from an audio time-series signal, and encodes the tone component signal and a residual time-series signal obtained by extracting a tone component signal from the audio time-series signal.
• the audio signal converting apparatus extracts a tone component signal from an audio time-series signal, encodes the tone component signal, and further extracts the tone component signal from the audio time-series signal.
• An audio signal decoding device for inputting a code sequence obtained by encoding a difference signal and decoding the code sequence, comprising: a code sequence decomposing means for decomposing the code sequence; Decoding the tone component time-series signal according to the obtained tone component information, and decoding the residual component time-series signal according to the residual component information obtained by the code string decomposing means. And a tone component time series signal obtained by the tone component decoding means and a residual component time series signal obtained by the residual component decoding means. Add the above sound An adding means for restoring a time-series signal.
• Such an audio signal decoding apparatus extracts a tone component signal from an audio time-series signal, and encodes the tone component signal and a residual time-series signal obtained by extracting a tone component signal from the audio time-series signal. Decodes the code sequence and restores the acoustic time-series signal.
• the recording medium according to the present invention is a computer-controllable recording medium on which an acoustic signal encoding program for encoding an acoustic time-series signal is recorded, wherein the acoustic signal encoding program comprises: A tone component encoding step of extracting and encoding a tone component signal; and a residual component for encoding the residual time-series signal obtained by extracting the tone component signal from the acoustic time-series signal in the tone component encoding process. And an audio signal encoding program characterized by having a difference component encoding step.
• Such a recording medium includes a sound component for extracting a tone component signal from an acoustic time-series signal, and encoding the tone component signal and a residual time-series signal obtained by extracting a tone component signal from the acoustic time-series signal.
• An encoding program is recorded. Further, the recording medium according to the present invention extracts a tone component signal from an acoustic time-series signal. Encodes the tone component signal, and further extracts a residual time-series obtained by extracting the tone component signal from the acoustic time-series signal.
• Such a recording medium includes a code obtained by extracting a tone component signal from an acoustic time-series signal, and encoding the tone component signal and a residual time-series signal obtained by extracting a tone component signal from the acoustic time-series signal.
• a sound signal decoding program that decodes the sequence and restores the sound time-series signal is recorded.
• FIG. 1A and FIG. 1B are diagrams for explaining a conventional method of extracting a tone component, FIG. 1A shows a spectrum before removing a tone component, and FIG. And shows the spectrum of the noise component after removing the tone component.
• FIG. 2 is a diagram illustrating a configuration of the audio signal encoding device according to the present embodiment.
• 3A to 3C are diagrams for explaining a method of smoothly connecting the extracted time-series signal to the preceding and succeeding frames.
• FIG. 3A shows a frame in MDCT
• FIG. Figure 3C shows the window function used for combining with the previous and next frames.
• FIG. 4 is a diagram illustrating a configuration of a tone component encoding unit of the acoustic signal encoding device.
• FIG. 5 is a diagram illustrating a first configuration of a tone component encoding unit that includes a quantization error in a residual time-series signal.
• FIG. 6 is a diagram illustrating a first configuration of a tone component encoding unit that includes a quantization error in a residual time-series signal.
• FIG. 7 is a diagram illustrating an example in which a normalization coefficient is determined based on the maximum amplitude values of a plurality of extracted sine waves.
• FIG. 8 is a flowchart showing a series of operations of the audio signal encoding device having the tone component encoding unit of FIG.
• FIGS. 9A and 9B are diagrams illustrating parameters of a pure sound waveform
• FIG. 9A shows an example using frequency and amplitude of sine wave and cosine wave
• FIG. 9B shows frequency, amplitude and An example using a phase will be described.
• FIG. 10 is a flowchart showing a series of operations of the audio signal encoding device having the tone component encoding unit of FIG.
• FIG. 11 is a diagram illustrating a configuration of an audio signal decoding device according to the present embodiment.
• FIG. 12 is a diagram illustrating a configuration of a tone component decoding unit of the acoustic signal decoding device.
• FIG. 13 is a flowchart illustrating a series of operations of the acoustic signal decoding device.
• FIG. 14 is a diagram illustrating another configuration example of the residual component encoding unit of the acoustic signal encoding device.
• FIG. 15 is a diagram illustrating a configuration example of a residual signal decoding unit corresponding to the residual signal encoding unit in FIG.
• FIG. 16 is a diagram illustrating a second configuration example of the audio signal encoding device and the audio signal decoding device.
• FIG. 17 is a diagram illustrating a third configuration example of the acoustic signal encoding device and the acoustic signal decoding device.
• FIG. 2 shows an example of a configuration of the audio signal encoding device according to the present embodiment.
• the audio signal encoding apparatus 100 includes a tone / noise determination unit 110, a tone component encoding unit 120, a residual component encoding unit 130, a code It has a column generation unit 140 and a time series holding unit 150.
• the tone / noise determination unit 110 determines whether the input acoustic time-series signal S is a tone signal or a noise signal, and outputs a tone / noise determination code T / N according to the determination result. To switch the subsequent process.
• the tone component encoding section 120 extracts a tone component from the input signal and encodes the tone component signal.
• the tone / noise determining section 110 determines a tone from the input signal determined to be tonality.
• Component parameter N-TP is extracted.Tone component extractor 1 2 1 and tone component parameter N-TP obtained by tone component extractor 1 2 1 are normalized and quantized, and quantized tone component parameters. Normalization for outputting N-QTP • Quantization unit 122 is provided.
• the residual component encoding unit 130 outputs a residual component when the tone component signal is extracted by the tone component extracting unit 121 from the input signal determined to be tonality by the tone 'noise determining unit 110. It encodes the sequence signal RS or the input signal determined to be noise by the tone 'noise determination unit 110, and converts these time series signals into, for example, a Modified Discrete Cosine Transform (Modified Discrete Cosine Transform). (on: MDCT) etc. to convert the spectrum information NS into the spectrum information NS, and the spectrum information NS obtained by the spectrum conversion section 13 1 And a normalizing and quantizing unit 132 for outputting quantized and quantized spectral information QNS.
• Modified Discrete Cosine Transform Modified Discrete Cosine Transform
• the code sequence generator 140 includes a tone component encoder 120 and a residual component encoder 13 Generates and outputs a code sequence c based on information from 0.
• Time series holding section 150 holds the time series signal input to residual component encoding section 130. The processing in the time-series holding unit 150 will be described later.
• the acoustic signal encoding apparatus 100 provides a subsequent stage for each frame in accordance with whether the input acoustic time-series signal is a tone signal or a noise signal.
• Switch the encoding method In other words, the tone component signal is extracted using the method of general harmonic analysis (GHA) as described later, and its parameters are coded, and its parameters are encoded.
• GAA general harmonic analysis
• the residual signal from which the tone component signal has been extracted and the noise signal are encoded, for example, after a spectrum transform using MDCT.
• an analysis frame (encoding unit) of an MDCT generally used for a spectrum transform needs to overlap a preceding and succeeding analysis frame by 1.2 frames.
• the analysis frame of the general harmonic analysis in the tone component encoding process can also have a 1/2 frame overlap with the analysis frame before and after, so that the extracted time-series signal is smooth with the extraction time-series signal of the previous and next frames. It is possible to connect to
• the time series signal of section A at the time of analysis of the first frame and the time series signal of section A at the time of analysis of the second frame And must not be different. For this reason, in the residual component encoding process, it is necessary to complete the tone component extraction in the interval A at the time of the spectrum conversion of the first frame, and it is preferable to perform the following process. .
• pure tone analysis is performed by the general sum analysis in the section of the second frame shown in FIG. 3B.
• waveform extraction is performed based on the obtained parameters, and the extraction section is set to the section that overlaps the first frame.
• the pure tone analysis by the general harmonic analysis in the section of the first frame has already been completed, and the waveform extraction in this section is performed based on the parameters obtained in each of the first and second frames. Do it. If the first frame is determined to be a noise signal, waveform extraction is performed based only on the parameters obtained in the second frame.
• the extracted time-series signals extracted in each frame are synthesized as follows. That is, as shown in FIG.
• Equation (1) the time series signal based on the parameters analyzed in each frame is multiplied by a window function that adds 1, such as the Hamiing function shown in Equation (1), and A time-series signal that is smoothly spread from the first frame to the second frame is synthesized.
• L is the frame length, that is, the length of the coding unit.
• the synthesized time-series signal is extracted from the input signal.
• a residual time-series signal in the section where the first frame and the second frame overlap (overlap section) is obtained, and this residual time-series signal is used for the second half of the first frame.
• the residual component coding of the first frame is performed by the residual time series signal of the first frame by using the residual time series signal and the residual time series signal of the first half frame of the first frame already stored. This is performed by constructing a signal, performing a spectrum transform on the residual time-series signal in the first frame, and normalizing and quantizing the obtained spectrum information.
• the synthesis of the tone component and the synthesis of the residual component are performed in the same frame during decoding. Becomes possible.
• the window function described above is applied only to the extracted time-series signal extracted in the second frame. .
• the obtained time-series signal is extracted from the input signal, and the residual time-series signal is similarly used as the residual time-series signal of the latter half 1Z2 frame of the first frame.
• the audio signal encoding apparatus 100 includes a time-series holding section 150 before a residual component encoding section 130, as shown in FIG. It has a configuration.
• the time-series holding unit 150 holds residual time-series signals for every 1Z2 frames.
• the tone component encoding unit 120 has parameter holding units 21 L 5, 22 17, and 23 19, and outputs waveform parameters and extracted waveform information in the previous frame. I do.
• the tone component encoding unit 120 shown in FIG. 2 specifically has a configuration as shown in FIG.
• the general harmonic analysis proposed by Wiener is applied to frequency analysis, tone component synthesis, and extraction in tone component extraction.
• This method is an analysis method that extracts a sine wave with the minimum residual energy from the original time-series signal in the analysis block and repeats the same operation on the residual signal.
• Frequency components can be extracted one by one in the time domain.
• the frequency resolution can be set freely, and more detailed frequency analysis is possible compared to methods such as Fast Fourier Transformation (FFT) and MDCT.
• FFT Fast Fourier Transformation
• the tone component encoding unit 210 shown in FIG. 4 includes a tone component extraction unit 211 and a normalization / quantization unit 212.
• the tone component extraction unit 211 and the normalization / quantization unit 212 are the same as the tone component extraction unit 121 and the normalization / quantization unit 122 shown in FIG.
• the pure tone analysis unit 2111 analyzes a pure tone component in which the energy of the residual signal is minimized from the input acoustic time-series signal S, and obtains a pure tone waveform parameter.
• the TP is supplied to the pure tone synthesizing unit 211 and the parameter holding unit 211.
• the pure tone synthesizer 2 1 1 2 synthesizes the pure tone waveform time-series signal TS of the pure tone component analyzed by the pure tone analyzer 2 1 1 1, and is synthesized by the pure tone synthesizer 2 1 1 2 in the subtracter 2 1 1 3.
• the pure sound waveform time series signal TS is extracted from the input acoustic time series signal S.
• the end condition determination unit 2 1 1 4 is obtained by the pure tone extraction in the subtracter 2 1 1 3 It is determined whether or not the residual signal satisfies the termination condition of the tone component extraction, and the residual signal is used as the next input signal of the pure tone analyzer 2 1 1 1 until the termination condition is satisfied. Is switched so that is repeated. This termination condition will be described later.
• the parameter holding unit 2 1 15 holds the pure tone waveform parameter TP in the current frame and the pure tone waveform parameter PrevTP in the previous frame, and normalizes and supplies the pure tone waveform parameter PrevTP in the previous frame to the quantization unit 2 I 20. I do. Also, the pure waveform parameter TP in the current frame and the blunt waveform parameter PrevTP in the previous frame are supplied to the extracted waveform synthesis unit 211.
• the extracted waveform synthesizing unit 2 1 16 synthesizes the time-series signal based on the pure sound waveform parameter TP in the current frame and the time-series signal based on the pure sound waveform parameter PrevTP in the previous frame using, for example, the Hanning function described above. Generate the time-series tone component signal N-TS in the (overlap section).
• the tone component time-series signal N-TS is extracted from the input acoustic time-series signal S, and the residual time-series signal KS in the overlapping section is output.
• the residual time series signal RS is supplied to and held by the time series holding unit 150 in FIG. 2 described above.
• the normalization / quantization unit 2120 normalizes and quantizes the pure sound waveform parameter PrevTP in the previous frame supplied from the parameter holding unit 2115, and quantizes the tone component parameter PrevN in the previous frame. -Output QTP.
• a tone component encoding unit 2200 shown in FIG. 5 includes a normalization / quantization unit for normalizing and quantizing information of the tone signal. 2 2 12 is included in the tone component extraction section 2 210.
• the pure tone analyzing unit 2 211 analyzes a pure tone component in which the energy of the residual signal is minimized from the input acoustic time-series signal S, and generates a pure tone waveform parameter.
• TP is supplied to the normalization / quantization unit 2 2 1 2.
• the normalization / quantization unit 2 2 1 2 normalizes and quantizes the pure tone waveform parameter TP supplied from the pure tone analysis unit 2 2 1 1 and inversely quantizes the quantized pure tone waveform parameter QTP. It is supplied to the inverse normalization unit 222 and the parameter holding unit 222.
• the inverse quantization / inverse normalization unit 2 2 1 3 inversely quantizes and inverse normalizes the quantized pure tone waveform parameter QTP, and converts the inversely quantized pure tone waveform parameter TP 'to a pure tone synthesis unit 2 2 1
• the pure tone synthesizer 2 2 1 4 synthesizes the pure tone waveform time series signal TS of the pure tone component based on the dequantized pure tone waveform parameter TP ′, and the pure tone synthesizer 2 2
• the pure sound wave time series signal TS synthesized in 14 is extracted from the input acoustic time series signal S.
• the termination condition determination unit 2 2 16 determines whether the residual signal obtained by the pure tone extraction in the subtracter 2 2 15 satisfies the termination condition of the tone component extraction, and satisfies the termination condition. Up to this point, switching is performed so that the residual signal is used as the next input signal of the pure tone analyzer 2 2 1 1 and the pure tone extraction is repeated.
• Parameter holding unit 2 2 1 7 holds a pure sound waveform parameter TP 'that is pure sound waveform parameters QTP and inverse quantization are quantized and the quantized tone component parameters PrevN in the previous frame - Output QTP . Further, the dequantized pure sound waveform parameter TP 'in the current frame and the dequantized pure sound waveform parameter PrevTP' in the previous frame are supplied to the extracted waveform synthesizing unit 222.
• the extracted waveform synthesizing unit 2 2 18 forms the time series signal based on the pure sound waveform parameter TP 'in the dequantized current frame and the time series signal based on the pure sound waveform parameter PrevTP' in the dequantized previous frame. For example, combining is performed using the Hanning function described above, and a tone component time-series signal N-TS in an overlapping section (overlap section) is generated.
• the tone component time-series signal N-TS is extracted from the input acoustic time-series signal S, and a residual time-series signal RS in an overlapping section is output. This residual time-series signal RS is supplied to and held by the time-series holding unit 150 in FIG.
• the tone component information is also normalized and quantized in the tone component encoder 230 in FIG.
• the normalization / quantization unit 2 3 15 is included in the tone component extraction unit 2 310.
• the pure tone analysis unit 2 3 1 1 A pure tone component in which the energy of the residual signal is minimized is analyzed from the obtained acoustic time-series signal S, and a pure tone waveform parameter TP is supplied to the pure tone synthesizing section 2 3 12 and the normalization / quantization section 2 3 15.
• the pure tone synthesizing section 2 3 1 2 synthesizes the pure tone waveform time series signal TS of the pure tone component analyzed by the pure tone analyzing section 2 3 1 1, and is synthesized by the pure tone synthesizing section 2 3 1 2 in the subtracter 2 3 1 3.
• the pure sound waveform time series signal TS is extracted from the input acoustic time series signal S.
• the termination condition determination unit 2 3 1 4 determines whether the residual signal obtained by the pure tone extraction in the subtractor 2 3 1 3 satisfies the termination condition of the tone component extraction, and determines whether the termination condition is satisfied. Switching is performed so that the residual signal is used as the next input signal of the pure tone analyzer 2 3 1 1 and the pure tone extraction is repeated.
• the normalization / quantization unit 2 3 15 normalizes and quantizes the pure tone waveform parameter TP supplied from the pure tone analysis unit 2 3 1 1, and dequantizes the quantized pure tone waveform parameter N-QTP. ⁇ Supplied to the inverse normalizing section 2 3 16 and the parameter holding section 2 3 19.
• the inverse quantization and inverse normalization unit 2 3 1 6 inversely quantizes and inverse normalizes the quantized pure sound waveform parameter N-QTP, and holds the inversely quantized pure sound waveform parameter N-TP 'as a parameter holding unit. Supply to 2 3 1 9
• the parameter holding unit 2 3 19 holds the quantized pure tone waveform parameter N-QTP and the inversely quantized pure tone waveform parameter N-TP ', and quantizes the tone component parameter PrevN-QTP in the previous frame. Is output. Further, the dequantized pure sound waveform parameter N-TP 'in the current frame and the dequantized pure sound waveform parameter PrevN-TP' in the previous frame are supplied to the extracted waveform synthesizing unit 2317.
• the extracted waveform synthesizing unit 2 3 17 generates a time series signal based on the pure-sound waveform parameter N-TP 'in the dequantized current frame and a time series based on the pure-sound waveform parameter PrevN-TP' in the dequantized previous frame.
• the signal and the signal are synthesized using, for example, the above-mentioned Hanning function, and a tone component time-series signal N-TS in an overlapping section is generated.
• the subtractor 2 3 18 extracts the tone component time-series signal N-TS from the input acoustic time-series signal S, and outputs a residual time-series signal RS in an overlapping section.
• the residual time series signal RS is supplied to and held by the time series holding unit 150 in FIG.
• the normalization coefficient for the amplitude is fixed at a value equal to or larger than the maximum value that can be taken. For example, when an audio time-series signal recorded on a music compact disc (CD) is used as an input signal, quantization is performed using 96 dB as a normalization coefficient. Since the normalization coefficient is a fixed value, it need not be included in the code string.
• CD music compact disc
• step S1 an acoustic time series signal in a certain analysis section (the number of samples) is input.
• step S2 it is determined whether or not the input time-series signal is tonic in the analysis section.
• Various methods can be considered as a discrimination method.
• the input time-series signal X (t) is subjected to spectrum analysis by FFT or the like, and the average value AVE (X ( k)) and the maximum value Ma x (X
• step S2 if it is determined that the image is a tone, the process proceeds to step S3. If it is determined that the image is a noise, the process proceeds to step S10.
• step S3 a frequency component with the minimum residual energy is determined from the input time-series signal.
• the input time-series signal x The residual component when a pure sound waveform of frequency f is extracted from (t) is as shown in the following equation (3).
• L is the length of the analysis interval (the number of samples).
• step S4 the pure time waveform of the frequency f obtained in step S3 is converted into the input time-series signal x as in the following equation (7). (T).
• step S5 it is determined whether or not the extraction end condition is satisfied.
• the extraction termination conditions include, for example, that the residual time-series signal is not a tone signal, that the energy of the residual time-series signal has decreased by more than a predetermined value, and that the energy of the input time-series signal has decreased. And the fact that the amount of decrease in the residual time-series signal due to the extraction of the pure tone is equal to or less than the threshold.
• step S5 If the extraction termination condition is not satisfied in step S5, the process returns to step S3.
• the residual time series signal obtained by equation (7) is the next input time series signal X i
• step S5 The process from step S3 to step S5 is repeated N times until the extraction end condition is satisfied. If the extraction termination condition is satisfied in step S5, the process proceeds to step S6.
• step S6 normalization and quantization of the obtained N pieces of pure tone information, that is, tone component information N-TP, are performed.
• the pure tone information is the frequency f n , amplitude S fn , amplitude C fn of the extracted pure sound waveform as shown in Fig. 9A, or the frequency ⁇ » amplitude A fn , phase as shown in Fig. 9B.
• P can be considered.
• the frequency f garbage, the amplitude S fn , the amplitude C fn , the amplitude A fn , and the phase ⁇ ⁇ ⁇ are expressed by the following equations (8) to ( 10).
• step S7 the quantized tone component information N-QTP is dequantized and denormalized to obtain tone component information N-TP '.
• tone component information N-TP ' In this way, by normalizing and quantizing the tone component information once and then dequantizing and denormalizing, in the decoding process of the acoustic time series signal, there is no difference from the tone component time series signal extracted here. Time series signals can be added.
• step S8 for each of the tone component information PrevN-TP 'in the previous frame and the tone component information N-TP' in the current frame, as shown in the following equation (11), the tone component time series signal N -Generate TS.
• NTS (t) Y (S ' fi sin (2 f n t) + C' ft cosf2Kf n t >> (0 ⁇ t ⁇ L) (1 1)
• tone component time-series signals N-TS are combined in the overlapping section, and the tone component time-series signal N-TS in the overlapping section is obtained.
• step S9 as shown in the following equation (1 2), the synthesized tone component time series signal N-TS is subtracted from the input time series signal S, and the residual time series signal RS for 1Z2 frames is subtracted.
• RS (t) S (t)-NTS (t) ( ⁇ ⁇ t ⁇ L) (12)
• step SI 1 the residual time series signal RS for 12 frames or the 12 frames of the input signal determined to be noisy in step S 2 and the 1/2 frame already held
• One frame to be currently coded is constituted by the residual time series signal RS of the minute or the input signal of a half frame, and the spectrum is transformed by DFT or MDCT using this frame.
• step S11 normalization and quantization of the obtained spectrum information are performed.
• step S12 it is determined whether or not the quantization information QI such as quantization accuracy and quantization efficiency is consistent.
• the quantization accuracy of the pure sound waveform parameter is too high, and sufficient quantization accuracy cannot be secured for the spectrum information.For example, the quantization accuracy and the quantization of the pure sound waveform parameter and the spectrum information of the residual time series signal are not sufficient. If the conversion efficiencies are not consistent, the quantization accuracy of the pure sound waveform parameter is changed in step S13, and the process returns to step S6.
• step S12 If it is determined in step S12 that the quantization accuracy and the quantization efficiency are consistent, the process proceeds to step S14.
• step S14 a code string is generated in accordance with the obtained pure sound waveform parameters and the spectrum information of the input signal determined to be a residual time-series signal or noise, and in step S15, the code string is generated. Output a code string.
• the audio signal encoding apparatus By performing the above-described processing, the audio signal encoding apparatus according to the present embodiment extracts a tone component signal from an audio time-series signal in advance, and performs efficient extraction for the tone component and the residual component. Encoding can be performed.
• the tone component encoding unit 120 is configured as shown in FIG. Although the processing of the acoustic signal encoding apparatus 100 having the configuration has been described, the processing of the acoustic signal encoding apparatus 100 when the tone component encoding unit 120 has the configuration shown in FIG. As shown in the flowchart of FIG. 8, the tone component encoding unit 120 is configured as shown in FIG. Although the processing of the acoustic signal encoding apparatus 100 having the configuration has been described, the processing of the acoustic signal encoding apparatus 100 when the tone component encoding unit 120 has the configuration shown in FIG. As shown in the flowchart of FIG.
• step S 21 a time-series signal in a certain analysis section (the number of samples) is input.
• step S22 it is determined whether or not the input time-series signal has a tone characteristic in the analysis section.
• This determination method is the same as the method in FIG. 8 described above.
• step S23 a frequency f i at which the residual energy is minimized is obtained from the input time-series signal.
• step S24 normalization and quantization of the pure sound waveform parameter TP are performed.
• the pure sound waveform parameters may be the frequency f amplitude S fl , amplitude C, the frequency fi, the amplitude A fl , and the phase P fl of the extracted pure sound waveform.
• step S25 the quantized pure tone waveform parameter QTP is inversely quantized and denormalized to obtain a pure tone waveform parameter TP '.
• step S26 a pure sound waveform time-series signal TS to be extracted is generated according to the following equation (13) according to the pure sound waveform parameter TP '.
• step S27 the pure sound waveform of the frequency fi obtained in step S23 is converted to the input time-series signal X as in the following equation (14).
• (T). x 1 (t) x 0 (t)-TS (t) (14)
• step S28 it is determined whether or not an extraction end condition is satisfied. If the extraction termination condition is not satisfied in step S28, the process returns to step S23.
• the residual time-series signal obtained by Expression (10) is used as the next input time-series signal X i (t).
• the processing from step S23 to step S28 is repeated N times until the extraction end condition is satisfied. If the extraction termination condition is satisfied in step S28, the process proceeds to step S29.
• step S29 according to the pure sound waveform parameter PrevTP 'in the previous frame and the pure sound waveform parameter TP' in the current frame, a tone component time-series signal N-TS for 1/2 frame to be extracted is synthesized.
• step S30 the synthesized tone component time-series signal N-TS is subtracted from the input time-series signal S to obtain a half-frame residual time-series signal RS.
• step S31 the residual time-series signal RS for the 12 frames or the one-two frames of the input signal determined to be noisy in the step S22 is already held.
• One frame is composed of the residual time series signal RS for 1/2 frame or the input signal for 1/2 frame, and this is spectrally transformed by DFT or MDCT.
• step S32 normalization and quantization of the obtained spectrum information are performed.
• step S33 it is determined whether or not the quantization information QI such as quantization accuracy and quantization efficiency is consistent.
• the quantization accuracy of pure sound wave parameters is too high, and sufficient quantization accuracy cannot be secured for spectrum information. If the quantization accuracy and the quantization efficiency of the pure waveform parameter and the spectrum information of the residual time series signal are not consistent, change the quantization precision of the pure waveform parameter in step S34. Then, the process returns to step S23. If it is determined in step S33 that the quantization accuracy and the quantization efficiency are consistent, the process proceeds to step S35.
• step S35 a code sequence is generated in accordance with the obtained blunt sound waveform parameters and the spectrum information of the residual time series signal or the input signal determined to be noisy, and in step S36 , And outputs the code string.
• FIG. 11 shows a configuration of an audio signal decoding apparatus according to the present embodiment.
• the audio signal decoding apparatus 400 includes a code string decomposition section 410, a tone component decoding section 420, a residual component decoding section 4330, and an addition. Vessel 44 0.
• the code sequence decomposing unit 410 decomposes the input code sequence into tone component information N-QTP and residual component information QNS.
• the tone component decoding unit 420 generates the tone component time-series signal N-TS 'according to the tone component information N-QTP, and the tone component information N-QTP obtained by the code sequence decomposition unit 410.
• Inverse quantization and inverse normalization unit 4 2 1 for inverse quantization and inverse normalization of tone component time series signal according to tone component parameter N-TP 'obtained by inverse quantization and inverse normalization unit 4 2 1
• a tone component synthesizing section 422 that synthesizes and outputs N-TS '.
• the residual component decoding section 4300 generates the residual time-series signal RS 'according to the residual component information QNS, and converts the residual component information QNS obtained by the code sequence
• the inverse quantization and inverse normalization unit 431, which performs inverse quantization and inverse normalization, and the spectrum information NS 'obtained by the inverse quantization and inverse normalization unit 431 are used as the inverse spectrum.
• an inverse spectrum conversion unit 432 for converting and generating a residual time-series signal RS ′.
• the adder 440 combines the output of the tone component decoding unit 420 with the output of the residual component decoding unit 430, and outputs a restored signal S ′.
• the audio signal decoding apparatus 400 in the present embodiment decomposes the input code string into tone component information and residual component information, and performs a decoding process according to each.
• the tone component decoding section 420 has a configuration as shown in FIG. I can do it.
• the tone component decoding section 500 has an inverse quantization * inverse normalization section 5110 and a tone component synthesis section 5200.
• the inverse quantization / inverse normalization unit 510 and the tone component synthesis unit 520 are the same as the inverse quantization / inverse normalization unit 421 and the tone component synthesis unit 422 shown in FIG. It is.
• the inverse quantization / inverse normalization unit 5100 inversely quantizes and inverse normalizes the input tone component parameter N-QTP to obtain the tone component parameter N.
• -Pure tone waveform parameters corresponding to each pure tone waveform of TP 'TP' 0, TP'1, , 5 2 1 1, ⁇ ⁇ ⁇ , and 52 IN.
• Pure tone synthesizer 5 2 1. , 5 2 1 1, ⁇ , 5 2 1 N are the pure tone waveform parameters TP '0, TP' 2, ⁇ , ⁇ ' ⁇ supplied from the inverse quantization and inverse normalization unit 5 10. Based on this, one pure sound waveform TS ′ 0, TS ′ 1,..., ⁇ 3 ′ ⁇ is synthesized and supplied to the adder 5 2.
• the pure tone synthesizer 5 2 1. , 5 2 1 1,..., 52 2
• the pure sound waveforms TS '0, TS' 1, ⁇ ' ⁇ , ⁇ 5' ⁇ supplied from IN are synthesized and output as a tone component time-series signal N-TS '. Power.
• step S41 the code sequence generated by the above-described audio signal coding apparatus 100 is input.
• step S42 the coded sequence is converted into tone component information and residual signal information. Decompose into
• step S43 it is determined whether or not a tone component parameter exists in the decomposed code string. If the tone component parameter exists, the process proceeds to step S44. If the tone component parameter does not exist, the process proceeds to step S46. In step S44, each parameter of the tone component is dequantized and denormalized to obtain each parameter of the tone component signal.
• step S45 the tone component waveform is synthesized according to each parameter obtained in step S44, and a tone component time series signal is generated.
• step S46 the residual signal information obtained in step S42 is inverse-quantized and inverse-normalized to obtain a spectrum of the residual time-series signal.
• step S47 the spectrum information obtained in the step S46 is inversely transformed, and a residual component time series signal is generated.
• step S48 the time-series signal of the tone component generated in step S45 and the time-series signal of the residual component generated in step S47 are added on a time series to obtain a restored time-series signal. Then, in step S49, the restored time-series signal is output.
• the audio signal decoding apparatus 400 in the present embodiment performs the above-described processing to restore the input audio time-series signal.
• step S43 it is determined whether or not a tone component parameter exists in the decomposed code string.However, without performing the determination, the process proceeds directly to step S44. You can do it. In this case, if there is no tone component parameter, in step S48, 0 is synthesized as a tone component time-series signal.
• the residual component encoding unit 130 shown in FIG. 2 may be replaced with one having the configuration shown in FIG.
• the residual component encoding unit 7100 includes a spectrum transforming unit 7101 that transforms the residual time-series signal RS into spectrum information RSP, It has a normalization unit 7102 that normalizes the spectrum information RSP obtained by the vector conversion unit 7101 and outputs the normalization information N. That is, the residual component encoding unit 7100 only normalizes the spectrum information and does not perform quantization, and outputs only the normalized information N to the decoding side.
• the decoding side has a configuration as shown in Fig. 15. That is, as shown in FIG. 15, the residual component decryption unit 7200 includes a random number generation unit 7201 that generates pseudo-spectrum information GSP using random numbers having an appropriate random number distribution, and a normal Inverse normalization unit 7202 that inversely normalizes pseudo-spectrum information GSP generated by random number generation unit 7201 according to quantization information, and inverse normalization by inverse normalization unit 7202
• the pseudo-spectrum information RSP ' is regarded as pseudo-spectrum information, and the inverse spectrum transform is performed to generate a pseudo residual time-series signal RS'.
• the random number distribution is close to the information distribution obtained when the general acoustic signal or noise signal is subjected to the spectrum transform and normalized. Good thing.
• a plurality of random number distributions are prepared, and the code By analyzing which distribution is optimal at the time of decoding, including the ID information of the optimal distribution in the code string, and generating random numbers using the random number distribution of the ID information referenced at the time of decoding, a more similar residual is obtained. It is possible to generate a difference time series signal.
• the encoded code sequence can be decoded by a method corresponding to the encoding side.
• the present invention is not limited to only the above-described embodiment.
• a configuration example of the audio signal encoding device and the audio signal decoding device for example, as shown in FIG.
• a configuration may be considered in which the acoustic time-series signal S is divided into a plurality of frequency bands, and each band is processed and encoded, and after decoding, the frequency bands are combined. The following is a brief description.
• the acoustic signal encoding device 8 10 is divided into a plurality of frequency bands by a band division filter unit 8 11 1 for dividing the input acoustic time series signal S into a plurality of frequency bands.
• Band signal encoding sections 812, 813, 814 for obtaining tone component information N-QTP and residual component information QNS from the input signal, and tone component information N-QTP and
• a code sequence generator 815 for generating a code sequence C from the QNS or residual component information QNS.
• the band signal encoders 8 12, 8 13, and 8 14 are composed of the above-described tone 'noise determiner, tone component encoder, and residual component encoder.
• the band signal encoding unit 814 may include only the residual component encoding unit.
• the audio signal decoding device fi820 receives the code sequence C generated by the audio signal encoding device 8110, and outputs tone component information N-QTP and residual component information of a plurality of frequency bands.
• the band signal decoding units 8 22, 8 23, and 8 24 are composed of the above-described tone component decoding unit, residual component decoding unit, and adder. As in the case of the side, in a high frequency band where there is often no tone component, it may be configured with only the residual component decoding unit.
• the encoding efficiency of a plurality of encoding methods is compared, and the encoding efficiency is improved.
• a configuration in which the code sequence C according to the encoding method is selected is also conceivable. The following is a brief description.
• the audio signal encoding apparatus 900 includes a first encoding unit 901, which encodes an input audio time-series signal S by a first encoding method, and an input audio time-series signal.
• a second encoding unit 905 that encodes S in the second encoding system; and an encoding efficiency determination unit 9 that determines the encoding efficiency of the first encoding system and the second encoding system.
• the first encoding unit 901 encodes a tone component of the acoustic time-series signal S, and a residual component output from the tone component encoding unit 902.
• a residual component encoding unit 903 that encodes the difference time series signal, and tone component information N-QTP obtained by the tone component encoding unit 902 and the residual component encoding unit 903.
• a code sequence generation unit 904 that generates a code sequence C from the residual component information QNS.
• the second encoding section 905 is composed of a spectrum conversion section 906 for converting an input time-series signal into spectrum information SP, and the spectrum conversion section 906.
• a normalization / quantization unit 907 for normalizing and quantizing the obtained spectrum information SP, and a quantized spectrum obtained by the normalization / quantization unit 907
• a code string generation unit 908 that generates a code string C from the information QSP.
• the coding efficiency determination unit 909 inputs the coding information CI of the code sequence C generated by the code sequence generation unit 904 and the code sequence generation unit 908. By this means, the coding efficiency of the first coding unit 901 and the coding efficiency of the second coding unit 905 are compared to select the code string C to be actually output, and the switch 9 1 Controls 0.
• the switch 910 switches the code string C to be output according to the switch code F supplied from the coding efficiency determination unit 909.
• the switch 9110 is configured to supply the code sequence to a first decoding unit 921 described later.
• the second encoder When the code string C of 905 is selected, switching is performed so that the code string C is supplied to a second decryption unit 926 described later.
• the audio signal decoding device 920 decodes the input code string C by the first decoding scheme, and converts the input code string C into the first code string. And a second decryption unit 926 for performing decoding by the second decoding method.
• the first decryption unit 9221 is obtained by a code decomposition unit 922 that decomposes the input code string C into tone component information and residual component information, and the code decomposition unit 9222 described above.
• a tone component decoding unit 923 that generates a tone component time-series signal from the obtained tone component information, and a residual component time-series signal from the residual component information obtained by the code decomposition unit 9222.
• the second decryption unit 926 includes a code decomposition unit 927 that obtains quantized spectrum information from the input code string C, and a code decomposition unit 927 that obtains the quantized spectrum information.
• Inverse quantization and inverse normalization section 928 for inverse quantization and inverse normalization of the obtained quantized vector information, and the inverse quantization and inverse normalization section 928
• An inverse spectrum transform unit 929 for inversely transforming the vector information to obtain a time-series signal.
• the input code sequence C is decrypted by the decoding method corresponding to the encoding method selected by the acoustic signal encoding device 900.
• the force S mainly obtained by performing the spectrum transformation using the MDCT is not limited to this, and may be FFT, DFT, DCT, or the like. Also, the overlap between frames is not limited to 1/2 frame.
• the recording medium is configured as hardware, but it is also possible to provide a recording medium in which a program according to the above-described encoding method and decoding method is recorded. Furthermore, it is also possible to provide a recording medium in which a code string obtained by this and a signal obtained by decoding the code string are recorded.
• INDUSTRIAL APPLICABILITY According to the present invention as described above, a tone component signal is extracted from an acoustic time-series signal.
• a residual time-series signal obtained by extracting a tone component signal from the tone component signal and the acoustic time-series signal ⁇ By encoding, it is possible to prevent the spectrum from being spread due to the tone component generated at the local frequency, thereby preventing the encoding efficiency from deteriorating.

Landscapes

• Engineering & Computer Science (AREA)
• Physics & Mathematics (AREA)
• Signal Processing (AREA)
• Computational Linguistics (AREA)
• Health & Medical Sciences (AREA)
• Audiology, Speech & Language Pathology (AREA)
• Human Computer Interaction (AREA)
• Acoustics & Sound (AREA)
• Multimedia (AREA)
• Spectroscopy & Molecular Physics (AREA)
• Compression, Expansion, Code Conversion, And Decoders (AREA)

Priority Applications (2)

Applications Claiming Priority (2)

Publications (1)

Family

ID=19022496

Family Applications (1)

Country Status (5)

Cited By (2)

Families Citing this family (20)

Citations (12)

Family Cites Families (11)

• 2001
  • 2001-06-15 JP JP2001182384A patent/JP4622164B2/ja not_active Expired - Lifetime
• 2002
  • 2002-06-11 US US10/362,007 patent/US7447640B2/en not_active Expired - Lifetime
  • 2002-06-11 KR KR1020037002141A patent/KR100922702B1/ko not_active Expired - Fee Related
  • 2002-06-11 WO PCT/JP2002/005809 patent/WO2002103682A1/ja not_active Ceased
  • 2002-06-11 CN CNB028025245A patent/CN1291375C/zh not_active Expired - Fee Related

Patent Citations (12)

Also Published As

Similar Documents

Legal Events