JP3348759B2 - Transform coding method and transform decoding method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

[0001] 1. Field of the Invention [0002] The present invention relates to a conversion encoding method and a conversion decoding method for a signal containing a pitch component, such as a tone signal or a voice signal.

[0002]

2. Description of the Related Art At present, as a method for encoding an audio signal such as a tone signal or a voice signal with high efficiency, the audio signal is divided into frames, which are called frames and have a fixed interval of about 5 to 50 ms. Time to signal-
The frequency domain signal obtained by performing the frequency conversion is divided into an envelope shape of the frequency characteristic (an outline of the frequency characteristic) and a residual signal obtained by flattening the frequency domain signal with the outline of the frequency characteristic.
It has been proposed to separate the two pieces of information and encode each of them.

As a specific method of such an encoding method, an adaptive spectral perceptual entropy encoding method (ASPEC, Adaptive Spectral Perceptual Entropy Codin) is known.
g), transform coding method using weighted vector quantization (TCW
VQ, Transform Coding with Weighted Vector Quantizati
on), and MPeg-Audio Layer 3 (MP
EG-Audio Layer 3) and the like have been proposed.

[0004] These technologies are described in K. Brande
nburg, J. Herre, JDJohnston etal: "ASPEC: Adaptive
spectral entropy coding of high quality music sign
als ", Proc. AES'91, T. Moriya, H. Suda:" An 8 Kbit / s
transform coder for noisychannels ", Proc.ICASSP'89
pp.196--199 and ISO / IEC standard IS-11172-3.

Here, in order to realize highly efficient encoding by these encoding methods, it is desirable that the residual signal has as flat a frequency characteristic as possible. For this reason, the adaptive spectral hearing control entropy coding method (AS
PEC) or MPeg-Audio Layer 3 (M
In PEG-Audio Layer 3), as shown in FIG. 5, a frequency domain signal is divided into several sub-bands, and a signal in each sub-band is divided by a value called a scaling factor representing the strength of the band. The frequency characteristics of the residual signal are flattened by normalization.

On the other hand, as a method of flattening a frequency domain signal with higher efficiency than these methods, there is a method using a linear prediction analysis as shown in FIG. In this method, a frequency characteristic is flattened by driving a linear prediction analysis filter with a linear prediction coefficient obtained by linearly predicting an input signal.
This method is a method used in the above-described transform coding method using weighted vector quantization (TCWVQ).

[0007] Regarding each related technology such as linear prediction analysis, discrete cosine transform (DCT), and modified discrete cosine transform (MDCT), Saito and Nakata, "Basics of speech information processing".
Chapter 6 of Ohmsha, KRRao, P. Yip, Translated by Yasuda and Fujiwara, "Image Coding Technology-DCT and Its International Standards", Chapter2, HSMalvar, "Signal Processing with L
apedTransforms, "Artech House, and ISO / IEC Standard IS-11172-3.

[0008]

However, these encoding methods only normalize the general outline of the frequency characteristics and efficiently remove microscopic irregularities in the frequency characteristics due to pitch components of musical sounds and voices. It cannot be removed well. Therefore, this is an obstacle, and it is difficult for the above-mentioned conventional encoding method to achieve high efficiency when encoding an audio signal having a strong pitch component.

The present invention has been made in view of the above-mentioned problems, and provides a transform coding method and a transform decoding method capable of efficiently coding an audio signal containing a pitch component. It is an object.

[0010]

According to the first aspect of the present invention,
Convert musical or audio signals into frames at fixed time intervals
Divide and perform time-frequency conversion on each frame to
Time generates area signals - a frequency conversion stage, frequency domain
Generates an approximate signal of the area signal and quantizes the approximate signal.
A rough calculation / quantization stage that outputs a quantized rough index
Floor and the generalized signal obtained by quantizing the frequency domain signal
And planarization generating a flattened signal by dividing by the previous
Detect and quantize pitch components from the flattened signal and quantize
Pitch encoding step to output pitch component index
If, dividing the pitch component obtained by quantizing from the flattened signal
Output the quantized flattened signal index of the flattened signal
It is characterized by having a flattened signal quantization step of force.

According to a second aspect of the present invention, in the first aspect of the invention, in the pitch encoding step, a basic
Quantized seeking frequency, the fundamental from the frequency domain signal
Frequency that is a natural number multiple of frequency or closest to this frequency
Extract frequency samples as pitch component samples
Quantized, the basic pitch of the quantized pitch thus obtained
Wave index and quantized pitch component sample index
Is output as a quantized pitch component index .

According to a third aspect of the present invention, in the second aspect of the invention, in the pitch encoding step, the frequency domain signal
To a frequency that is a natural number multiple of the fundamental frequency or
The nearest frequency sample and the consecutive
Quantizing the pitch component by extracting the number of samples as one unit
Is characterized by

The invention according to claim 4 is the invention according to claim 2 or 3.
In the described invention, in the pitch encoding step, the pitch component
It is characterized in that the minute samples are vector-quantized collectively or for each unit .

According to a fifth aspect of the present invention, there is provided a quantized flattened signal.
Flattening signal regeneration that reproduces the flattening signal from the index
Step and pitch component from quantized pitch component index
Playback pitch and quantization rough index
And envelope signal reproducing step of reproducing the outline signal from the flat
The signal obtained by adding the pitch component to the digitized signal is
De-flattening step to reproduce the frequency domain signal by de-flattening
And performing a time-frequency inverse transform on the frequency domain signal.
Time to generate a tone signal or voice signal in the time domain-
It is characterized by having a frequency inverse transform stage.

According to a sixth aspect of the present invention, in the fifth aspect of the invention, in the pitch reproducing step, the basic circumference of the pitch component is
The wave number is decoded, and the quantized pitch component
The pitch component sample decoded from the sample index
A frequency that is a natural number multiple of the fundamental frequency or this frequency
It is characterized by being arranged as a frequency domain signal at the closest frequency .

According to a seventh aspect of the present invention, in the invention according to the sixth aspect, at the pitch reproduction stage, a quantity is used as a pitch component.
The pitch decoded from the child pitch component sample index
Multiplied by a natural number multiple of the fundamental frequency or
Include the sample at the frequency closest to this frequency and
Frequency domain signal with multiple consecutive samples as one unit.
It is characterized by being arranged as a number .

The invention according to claim 8 is the invention according to claim 6 or 7.
In the described invention, in the pitch reproduction stage, the pitch component
It is characterized in that samples are decoded at once or for each unit .

[0018]

The musical tone or voice has a pitch, that is, a high / low pitch. The frequency domain signal obtained by frequency-converting this musical tone or voice contains pitch components arranged at a constant frequency interval. Therefore, the above-mentioned pitch component is also included in the residual signal obtained by normalizing the frequency domain signal with the outline of its own frequency characteristic. Since this pitch component appears as a spike having a large energy with respect to the entire power, the flatness of the residual signal is reduced to deteriorate the quantization efficiency. However, the present invention focuses on the fact that pitch components are arranged at regular intervals on the frequency axis, and subtracts the pitch components from the residual signal, thereby increasing the flatness of the residual coefficient with a small amount of additional information.

[0019]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described below with reference to the drawings. FIG. 1 is a diagram for explaining a transform coding method and a transform decoding method according to the present embodiment, where code A is an encoder, and B is a decoder. As shown, the encoder A includes a time-frequency converter 1, a global shape calculator / quantizer 2, a first flattener 3, a pitch encoder 4, an adder 5, and a fine spectrum shape calculator.・ Quantizer 6,
It comprises a second flattener 7 and a quantizer 8.

The time-frequency converter 1 is a time-domain input signal (music signal or audio signal such as a voice signal).
Is divided into frames at fixed time intervals, and time-frequency conversion is performed on each frame to generate a frequency domain signal.

FIG. 2 shows the frequency characteristics of the frequency domain signal. As shown in this figure, the frequency domain signal of the tone signal or the audio signal has a constant frequency interval p.
The pitch components arranged in the order are included. As a conversion method, Discrete Cosine Transf
ormation, DCT) and modified discrete cosine transform (Modified Dis
crete Cosine Transformation, MDCT).

The global shape calculator / quantizer 2 generates and quantizes a signal indicating a global shape of the frequency domain signal output from the time-frequency converter 1. Then, this signal is output to the first flattener 3 and
Output to the outside as a quantized global shape index.
As a method of calculating the general outline, a linear prediction spectrum or a scale factor that divides a frequency domain signal into a plurality of sub-bands and uses a representative value of each band to express an outline of the entire frequency domain signal may be used. Good.

When quantizing the linear prediction spectrum, the linear prediction parameters are converted into LSP parameters and quantized. Alternatively, it is converted into a K parameter and quantized.

The first flattener 3 calculates a global outline of the frequency domain signal output from the time-frequency converter 1.
The signal is flattened by being divided by the globally generalized signal output from the quantizer 2, and a first flattened signal is output.

Next, the pitch encoder 4 detects and encodes a pitch component from the first flattened signal. Also,
FIG. 3 is a diagram showing details of the pitch encoder 4. The first flattened signal is a pitch fundamental frequency extractor 4 shown in FIG.
a and pitch sample extractor 4b.

The pitch fundamental frequency extractor 4a has a first
The fundamental frequency of the pitch component (pitch fundamental frequency) is obtained by analyzing the flattened signal of That is, the pitch fundamental frequency extractor 4a calculates the cepstrum of the first flattening coefficient, and sets the maximum value as the fundamental period of the pitch component. Then, the pitch fundamental frequency is obtained by calculating the reciprocal of the fundamental period, and the pitch fundamental frequency quantizer 4c
Output to

In order to make the pitch fundamental frequency more accurate, the fundamental frequency at which the sum of the powers of the samples of the first flattened signal for each pitch fundamental frequency becomes maximum before and after the obtained pitch fundamental frequency is determined. A search may be made, and this may be newly set as the pitch fundamental frequency.

The pitch fundamental frequency quantizer 4c quantizes the pitch fundamental frequency thus obtained. That is, the pitch fundamental frequency quantizer 4c scalar-quantizes the logarithmic value of the pitch fundamental frequency, outputs the quantized pitch fundamental frequency index to the outside, and outputs the scalar-quantized signal to the pitch sample extractor 4b. Output to

The pitch sample extractor 4b converts the first flattened signal inputted from the first flattener 3 into a natural number of the quantized pitch fundamental frequency inputted from the pitch fundamental frequency quantizer 4c. One sample before and after the sample closest to the double frequency is extracted, and one set of these three samples is output to the pitch sample quantizer 4d as one pitch component sample group. Note that the number of the pitch component sample groups may be a fixed value or may be variable.

The pitch sample quantizer 4d quantizes the sample group of pitch components and outputs the quantized pitch component index to the outside, and decodes the quantized pitch component index to the adder. 5 is output. The quantization of the sample group is
Scalar quantization may be used, or vector quantization may be performed for each sample group including three samples. Further, all the sample groups may be collectively vector-quantized. The above is the processing performed in the pitch encoder 4.

Next, the adder 5 uses the quantized signal of the pitch component input from the pitch encoder 4 to calculate the pitch component from the first flattened signal input from the first flattener 3. A second flattened signal is generated by subtracting only the second flattened signal, and output to the fine spectrum rough shape calculator / quantizer 6 and the second flattener 7.

FIG. 4 is a diagram showing the frequency characteristics of the second flattened signal. As can be seen from a comparison with FIG. 2, the second flattened signal is obtained by removing the pitch component from the frequency domain signal output from the time-frequency converter 1.

The fine spectrum shape calculator / quantizer 6 comprises:
A fine spectral outline (fine spectral outline) is calculated from the second flattened signal and quantized. Then, the quantized signal is output to the outside as a quantized fine spectrum rough shape index and to the second flattener 7.

The fine spectral outline may be obtained by directly quantizing the fine spectral outline or by linearly synthesizing the fine spectral outline of a past frame.
Alternatively, the information of the quantized fine spectral outline of the past and current frames may be obtained by linear synthesis. Further, the fine spectrum outline may be obtained by convolving the absolute value of the second flattened signal with a window function having a width of about 3 to 5 or using the second flattened signal obtained by subband division. A representative value of the amplitude of the digitized signal may be prepared for each band, and this may be used as an outline.

The second flattener 7 divides the second flattened signal input from the adder 5 by the fine spectral outline obtained by the fine spectral outline calculator / quantizer 6 to flatten it. The signal is output to the quantizer 8 as a third flattened signal. The quantizer 8 performs scalar quantization or vector quantization on the third flattened signal, and outputs it as a quantization index to the outside.

In the case of vector quantization, all samples of a frame may be quantized collectively. However, it is better to divide the sample sequence of a frame into a plurality of subvectors and quantize each subvector. It is realistic in terms of computational complexity. Further, the division method may be a simple subband division or an interleave division in which samples are interleaved and then divided. Also, adaptive bit allocation may be performed in accordance with the amount of information necessary for quantization.

Next, the decoder B will be described. As shown in FIG. 1, the decoder B includes a regenerator 9, a fine spectrum rough shape regenerator 10, a first inverse flattener 11, and a pitch regenerator 1.
2, an adder 13, a global outline regenerator 14, a second inverse flattener 15, and an inverse time-frequency converter 16.

The regenerator 9 regenerates the third flattened signal from the quantization index transmitted from the encoder A. The regenerator 9 reproduces the third flattened signal by performing the inverse processing of the quantizer 8 and outputs the third flattened signal to the first inverse flattener 11. The fine-spectrum rough shape regenerator 10 recovers the fine-spectrum rough shape from the fine-spectrum rough shape quantization index transmitted from the encoder A.

The first inverse flattener 11 adds a fine spectral outline to the third flattened signal input from the regenerator 9 to regenerate the second flattened signal and sends the second flattened signal to the adder 13. Output. The pitch reproducer 12 reproduces the pitch component from the quantized pitch component index and the quantized pitch fundamental frequency index transmitted from the encoder A, and outputs the reproduced pitch component to the adder 13.

The adder 13 reproduces the first flattened signal by adding the pitch component inputted from the pitch reproducer 12 to the second flattened signal inputted from the first inverse flattener 11. , To the second inverse flattener 15. Further, the global shape regenerator 14 regenerates the global shape from the quantized global shape index transmitted from the encoder A, and outputs it to the second inverse flattener 15.

The second inverse flattener 15 adds the global shape input from the global shape regenerator 14 to the first flattened signal input from the adder 13, and Generate Then, the inverse time-frequency converter 16 performs inverse time-frequency conversion on the frequency-domain signal input from the second inverse flattener 15 and decodes the signal, and outputs a time-domain audio signal or tone signal.

[0042]

As described above, according to the present invention,
When encoding a tone signal or voice signal having a pitch component, the signal is efficiently encoded by utilizing the regularity of spike-like pitch components appearing in a frequency domain signal obtained by converting the signal into a frequency domain. . Therefore, a more flattened residual coefficient can be obtained, and the efficiency of the entire encoder can be increased.

[Brief description of the drawings]

FIG. 1 is a diagram illustrating an encoder and a decoder according to an embodiment of the present invention.

FIG. 2 is a diagram showing a frequency characteristic of an output signal of a time-frequency converter in the present invention.

FIG. 3 is a diagram showing a detailed configuration of a pitch encoder in the present invention.

FIG. 4 is a diagram showing a frequency characteristic of a second flattened signal in the present invention.

FIG. 5 is a first diagram illustrating a conventional transform encoding method.

FIG. 6 is a second diagram illustrating a conventional transform encoding method.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Time-frequency converter 2 Global shape calculation / quantizer 3 First flattener 4 Pitch encoder 5, 13 Adder 6 Fine spectrum shape calculation / quantizer 7 Second flattener 8 Quantum Modifier 9 regenerator 10 fine spectrum rough shape regenerator 11 first inverse flattener 12 pitch regenerator 14 global shape regenerator 15 second inverse flattener 16 time-frequency inverse converter

──────────────────────────────────────────────────続 き Continuation of the front page (56) References JP-A-1-239597 (JP, A) JP-A-57-62096 (JP, A) JP-A-57-161795 (JP, A) JP-A 63-209 37400 (JP, A) JP-A-7-261800 (JP, A) (58) Fields investigated (Int. Cl. ⁷ , DB name) G10L 11/00-11/04 G10L 19/00-19/02 G11B 20/10 H03M 7/30

Claims (8)

(57) [Claims] 1. A music signal or a voice signal for a certain period of time.
Divide into separate frames and convert each frame to time-frequency
Time-frequency conversion stage for generating a frequency domain signal
And generate an approximate signal of the frequency domain signal, and quantify the approximate signal.
Approximation calculation that outputs a quantized outline index
Quantizing the frequency domain signal with the quantized generalized signal;
A flattening step of generating a flattened signal by dividing, and detecting and quantizing a pitch component from the flattened signal to perform quantization.
Pitch encoding step that outputs a normalized pitch component index
If, removing the pitch component obtained by quantizing from the flattened signal
Output the quantized flattened signal index of the flattened signal
Transform coding method characterized by having a flattened signal quantization step that. 2. In the pitch encoding step, a fundamental frequency of a pitch component is obtained and quantized, and a frequency which is a natural number multiple of the fundamental frequency is obtained from a frequency domain signal.
Or pitch the sample at the frequency closest to this frequency
It is extracted as a component sample, quantized, and the quantized pitch fundamental frequency index obtained in this way is obtained.
And the quantized pitch component sample index
2. The transform encoding method according to claim 1 , wherein the output is output as a child pitch component index . 3. In the pitch encoding step, a frequency which is a natural number multiple of the fundamental frequency is obtained from the frequency domain signal.
Or the sample at the frequency closest to this frequency and
The pitch formation is performed with multiple consecutive samples including
3. The method according to claim 2 , wherein the component is extracted and quantized . 4. A pitch encoding step, comprising the steps of :
4. The transform coding method according to claim 2, wherein the pull is vector-quantized collectively or for each unit . 5. Flattening from a quantized flattened signal index
A flattening signal reproducing step of reproducing the signal, to reproduce a pitch component from the quantization pitch component index
Pitch reproduction step, and a rough signal for reproducing a rough signal from the quantized rough index.
Signal reproduction step, and the signal obtained by adding the pitch component to the flattened signal.
Inverse flattening to reproduce frequency domain signal by inverse flattening with shape signal
And performing time-frequency inverse transform on the frequency domain signal.
Time-frequency to generate a musical tone signal or audio signal in the area
Transform decoding method characterized by having a number inverse transformation stage. 6. A pitch reproducing step, wherein a fundamental frequency of a pitch component is decoded, and a quantized pitch component sample index is used as a pitch component.
Pitch component samples decoded from the
To a frequency that is a natural number times or the frequency closest to this frequency
The transform decoding method according to claim 5, wherein the signal is arranged as a frequency domain signal . 7. In the pitch reproduction stage, a pitch component
The pitch decoded from the quantized pitch component sample index
Switch component samples at natural frequency times the fundamental frequency or
Is the frequency sample closest to this frequency and contains
Frequency domain using multiple consecutive samples as one unit
7. The method according to claim 6, wherein the signal is arranged as a signal . 8. A pitch component sampling step, wherein a pitch component sample is sampled.
8. The conversion decoding method according to claim 6 , wherein the decoding is performed in a batch or for each unit .