JP4438654B2 - Encoding device, decoding device, encoding method, and decoding method - Google Patents

Description

  The present invention relates to an encoding device that encodes a signal, a decoding device that decodes an encoded signal, an encoding method, and a decoding method.

Conventionally, audio signal compression methods include μ-law, ADPCM (Adaptive Differential Pulse Code Modulation), MP3 (MPEG Audio Layer-3) used in music, VSELP (Vector Sum Excited Linear) used in mobile phones, etc. Prediction), G. A CELP (Code-Excited Linear Prediction) type compression method such as 729 has been put into practical use. Patent Document 1 discloses a technique using vector quantization as an audio compression technique.
Japanese Patent Laid-Open No. 10-63299

  When recording conversation or the like in language learning, a sampling frequency of about 16 kHz is considered to be an appropriate frequency that does not require a large amount of data while maintaining the characteristics of each language. However, there is a problem that the compression noise that appears in the CELP compression method is not appropriate for language learning. In addition, although μ-law and ADPCM have sufficient sound quality, since the encoding rate is high, there is a problem that recording time is shortened when these compression methods are used in a portable device. MP3 is mainly intended for compression of high-quality audio, and there is a problem that compression cannot be performed effectively at a sampling frequency of about 16 kHz.

  An object of the present invention is to improve the coding efficiency of a signal such as speech and realize effective coding.

In order to solve the above-described problem, an encoding apparatus according to claim 1 includes a framing unit that divides an input signal into frames, and performs frequency conversion on the input signal for each unit constituting one frame. Search and coefficient calculating means for calculating the frequency transform coefficients, the most similar frequency transform coefficients to the first frequency conversion coefficients of one unit in one frame, from the other frequency conversion coefficients of one unit in the same frame A first search unit that performs the search, a frequency conversion coefficient for one unit that is most similar searched by the first search unit, and number information that indicates the number of the frequency conversion coefficient for the unit in the frame. A first encoding means for converting, a specifying means for sequentially specifying one unit of frequency conversion coefficients for processing, and a frequency conversion coefficient for one unit specified by the specifying means. Also similar frequency transform coefficients, the previously second retrieval means for retrieving from the frequency transform coefficients of one unit which is coded, frequency conversion coefficients of one unit to be the processing target, the second Dividing means for dividing by the frequency conversion coefficient for one unit most similar searched by the searching means, and the result of division by the dividing means and the number information of the frequency conversion coefficient for one unit used for division in a predetermined encoding method A second encoding means for encoding, and the designation means, the second search means, the dividing means, and the coding until encoding of all frequency transform coefficients for one unit constituting one frame is completed. And control means for repeatedly executing the processing of the second encoding means .

According to a second aspect of the present invention, in the encoding device according to the first aspect, an error is calculated between the predetermined one unit of the frequency conversion coefficient and another one unit of the frequency conversion coefficient in the same frame. The first and second detecting means detect the most similar frequency conversion coefficient based on the error calculated by the error calculating means.

A decoding device according to claim 3 is a decoding device that decodes a signal encoded by the encoding device according to claim 1, and is encoded by a first encoding means of the encoding device. First decoding means for decoding the received signal to obtain a frequency transform coefficient for one unit and number information indicating the number of the frequency transform coefficient for one unit in the frame; Obtained by the second decoding means and the second decoding means for decoding the signal encoded by the encoding means and obtaining the division result and number information of the frequency conversion coefficient for one unit used for the division. Multiplication means for multiplying the division result by a frequency transformation coefficient used as a division value at the time of encoding specified by the number information, and inverse transformation for performing frequency inverse transformation on the signal obtained by the multiplication means It is characterized in that it comprises a means,

5. The encoding method according to claim 4, wherein a framing step for dividing the input signal into frames and a coefficient for calculating a frequency conversion coefficient by performing frequency conversion on the input signal for each unit constituting one frame. A calculation step, and a first search step for searching for a frequency conversion coefficient most similar to the frequency conversion coefficient for the first unit in one frame from the frequency conversion coefficients for other one unit in the same frame; First encoding that encodes the most similar frequency transform coefficient for one unit retrieved in the first retrieval step and number information indicating the number of the frequency transform coefficient for the one unit in the frame A step for specifying a frequency conversion coefficient for one unit to be processed in order in a frame, and a frequency conversion unit for one unit specified in the specifying step. A second search step for searching for a frequency conversion coefficient most similar to the frequency conversion coefficient for one unit already encoded, and a frequency conversion coefficient for one unit to be processed as the second conversion step. A division step of dividing by the most similar frequency conversion coefficient for one unit searched in the search step, a division result by the division step, and number information of the frequency conversion coefficient for one unit used for the division A second encoding step that encodes in accordance with a scheme, until the encoding of frequency conversion coefficients for all one unit constituting one frame is completed, the specifying step, the second search step, The division step and the processing of the second encoding step are repeatedly executed .

The encoding method according to claim 5 , further comprising an error calculating step of calculating an error between the predetermined one unit of frequency conversion coefficient and another one unit of frequency conversion coefficient in the same frame. In the first and second detection steps, the most similar frequency conversion coefficient is detected based on the error calculated by the error calculation means .

A decoding method according to claim 6 is a decoding method for decoding a signal encoded by the encoding method according to claim 4, and is encoded in a first encoding step of the encoding method. A first decoding step of decoding the received signal to obtain a frequency conversion coefficient for one unit and number information indicating the number of the frequency conversion coefficient for one unit in the frame; and a second of the encoding method Obtained by the second decoding step and the second decoding step to obtain the division result and number information of the frequency conversion coefficient for one unit used for the division. A multiplication step for multiplying the division result by a frequency transformation coefficient used as a division value at the time of encoding specified by the number information, and an inverse transformation for performing frequency inverse transformation on the signal obtained by the multiplication step Step It is characterized in that it consists of.

  According to the present invention, it is possible to improve encoding efficiency by dividing the frequency conversion coefficient for each unit in one frame by another frequency conversion coefficient most similar to each unit.

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

  FIG. 1A shows a configuration of an encoding apparatus 100 according to the embodiment of the present invention. The encoding device 100 includes a framing unit 1, a frequency converting unit 2, a dividing unit 3, and an encoding unit 4, as shown in FIG.

  The framing unit 1 divides an input signal such as an audio signal into frames of a certain length that is an encoding (compression) processing unit, and outputs the frame to the frequency conversion unit 2. One frame is composed of a plurality of units for performing frequency conversion.

  The frequency conversion unit 2 performs frequency conversion on the signal input from the framing unit 1 for each unit constituting the frame, and outputs the result to the division unit 3. The frequency conversion applied by the frequency conversion unit 2 is selected according to the characteristics of the input signal, such as FFT (Fast Fourier Transform) and MDCT (Modified Discrete Cosine Transform). For example, when the input signal is an audio signal, MDCT is often selected. Hereinafter, the frequency conversion coefficient calculated by one frequency conversion is referred to as “one unit of frequency conversion coefficient”.

The division unit 3 divides the frequency conversion coefficient for each unit in one frame by another frequency conversion coefficient most similar to each unit, and outputs the division result to the encoding unit 4. Hereinafter, the division process in the division unit 3 will be described in detail. In the following, there are n frequency conversion coefficients {C _i | i = 1,..., N} for one unit in one frame, and the frequency conversion coefficients for each unit are m frequency conversion coefficients C. Let _i = {c _ij | j = 1,..., m}.

式（１）で算出されるｅ_1kの値が最も小さくなる周波数変換係数Ｃ_kが、Ｃ₁に最も類似した周波数変換係数として得られる。Ｃ₁に最も類似した周波数変換係数Ｃ_kには除算処理が施されず、そのまま符号化部４で符号化される。 First, a frequency conversion coefficient most similar to the frequency conversion coefficient C ₁ for the first (temporally first) unit in the frame is searched from C _{2 to} C _n . As an index for searching for “the most similar frequency conversion coefficient”, an error from the frequency conversion coefficient for another unit in the same frame can be used. For example, the sum of square errors with C ₁ is calculated, and the frequency conversion coefficient having the smallest total value is set as “the frequency conversion coefficient most similar to C ₁ ”. A total e _1k of square errors of C ₁ and C _k (k ≠ 1) is calculated as shown in Expression (1).

The frequency conversion coefficient C _k that _minimizes the value of e _1k calculated by Expression (1) is obtained as the frequency conversion coefficient most similar to C ₁ . The frequency conversion coefficient C _k most similar to C ₁ is not subjected to division processing and is directly encoded by the encoding unit 4.

Next, the encoded frequency transform coefficient obtained by the above search is set as a “processed” frequency transform coefficient, and the following processing is performed until all frequency transform coefficients in one frame are “processed” in order from C _1. Repeat the division process as follows. Assuming that the division processing target is C _i , the most similar frequency conversion coefficient is searched from the frequency conversion coefficients that have been processed so far, and the division processing target C _i is the frequency conversion coefficient obtained by the search. Divide by. In order to detect the most similar frequency conversion coefficient, the total value of the square errors between the frequency conversion coefficients may be calculated in the same manner as in Equation (1). If the frequency conversion coefficient most similar to the frequency conversion coefficient C _i to be divided is C _j (i ≠ j), the division result C ′ _i = {c ′ _il | l = 1,... It is expressed as (2).

The encoding unit 4 performs an encoding process on the signal input from the division unit 3 using a predetermined encoding method, and outputs the encoded signal. The encoding unit 4 encodes the coefficient value C _k and its coefficient number k for the frequency conversion coefficient C _k most similar to C ₁ , and the division result C ′ for the remaining frequency conversion coefficients. _i and the coefficient number of the frequency conversion coefficient used as the division value are encoded. Here, as the encoding method, Huffman encoding, Range Coder encoding, or the like can be used.

  FIG. 1B shows a configuration of a decoding device 200 according to the embodiment of the present invention. The decoding device 200 is a device for decoding the signal encoded by the encoding device 100, and as illustrated in FIG. 1B, the decoding unit 5, the multiplication unit 6, the frequency inverse conversion unit 7, and the frame synthesis. It is comprised by the part 8.

The decoding unit 5 decodes the input encoded signal and outputs the decoded signal to the multiplication unit 6. Specifically, for the frequency transform coefficient C _k (frequency transform coefficient most similar to C ₁ ) directly encoded without division, the decoding unit 5 encodes the encoded coefficient value C _k and its coefficient number k. For the remaining frequency transform coefficients that have undergone the division process, the encoded division result C ′ _i and the coefficient number of the frequency transform coefficient used as the division value are decoded.

The multiplication unit 6 multiplies the decoded division result input from the decoding unit 5 by a frequency conversion coefficient that is a division value at the time of encoding, and outputs the multiplication result to the frequency inverse conversion unit 7. Specifically, the multiplying unit 6 converts the frequency transform coefficient that has been decoded and multiplied (the decoded frequency transform coefficient that is most similar to C ₁ and the other frequency transform coefficient that has been multiplied) to the “processed” frequency transform. Multiplication processing is repeated as follows until all frequency conversion coefficients in one frame become “processed” in order from C ₁ . Assuming that the decoded frequency conversion coefficient to be multiplied is C ′ _i = {c ′ _il | l = 1,..., M}, among the frequency conversion coefficients that have been processed so far, together with C ′ _i The frequency conversion coefficient of the decoded coefficient number j (frequency conversion coefficient used as a division value at the time of encoding) C ″ _j (i ≠ j) is multiplied by C ′ _i . The multiplication result C ″ _i = {c ” _il | l = 1,..., m} is calculated as shown in Equation (3).

  The frequency inverse transformation unit 7 performs frequency inverse transformation on the signal input from the multiplication unit 6 and outputs the result to the frame synthesis unit 8. The frame synthesizing unit 8 synthesizes frames that are processing units of encoding and decoding, and outputs the synthesized signal as a reproduction signal.

Next, the operation in this embodiment will be described.
First, with reference to the flowchart of FIG. 2, coefficient division processing executed in the division unit 3 of the encoding device 100 will be described.

First, a frequency conversion coefficient that is most similar to the first (temporally first) unit frequency conversion coefficient C ₁ in the frame is searched from the remaining frequency conversion coefficients in the frame (Step S1). S1). Next, the counter value i for designating the coefficient number of the frequency conversion coefficient is set to 1 (step S2).

Next, it is determined whether or not all frequency conversion coefficients within one frame have been processed (step S3). If it is determined in step S3 that all frequency conversion coefficients in one frame have not been processed (step S3; NO), it is determined whether or not the frequency conversion coefficient C _i with coefficient number _i has been processed. (Step S4).

If it is determined in step S4 that the frequency conversion coefficient C _i has not been processed (step S4; NO), the frequency conversion coefficient most similar to C _i is searched from the processed frequency conversion coefficients (step S5). Then, C _i is divided by the frequency conversion coefficient obtained by the search (step S6).

After the division process in step S6 or when it is determined in step S4 that the frequency conversion coefficient C _i has been processed (step S4; YES), the counter value i is incremented (step S7), and the current counter value i The process of steps S3 to S6 is repeated. In step S3, when it is determined that all the frequency conversion coefficients in one frame have been processed (step S3; YES), the coefficient division process ends.

  Next, an encoding process executed in the encoding unit 4 of the encoding device 100 will be described with reference to the flowchart of FIG.

First, the value of the frequency conversion coefficient most similar to the frequency conversion coefficient C ₁ for the first (first in time) unit in the frame and its coefficient number are encoded (step S10). Next, the division result obtained in step S6 of the coefficient division process and the coefficient number of the frequency conversion coefficient used as the division value are encoded (step S11), and the present encoding process ends.

  Next, a decoding process executed in the decoding unit 5 of the decoding device 200 will be described with reference to the flowchart of FIG.

  First, the value of the frequency transform coefficient and the coefficient number that are directly encoded without performing the division process on the encoding apparatus 100 side are decoded (step T1). Next, the encoded division result and the coefficient number of the frequency transform coefficient used as the division value are decoded (step T2), and this decoding process ends.

  Next, coefficient multiplication processing executed in the multiplication unit 6 of the decoding device 200 will be described with reference to the flowchart of FIG.

  First, the counter value i for designating the coefficient number of the frequency conversion coefficient is set to 1 (step T10). Next, it is determined whether or not all frequency conversion coefficients in one frame have been processed (step T11).

If it is determined in step T11 that all the frequency conversion coefficients in one frame have not been processed (step T11; NO), it is determined whether or not the division result C ′ _i of coefficient number _i has been processed. (Step T12).

If it is determined in step T12 that the division result C ′ _i has not been processed (step T12; NO), the frequency conversion of the coefficient number j decoded together with the division result C ′ _i from the processed frequency conversion coefficients. A coefficient (frequency conversion coefficient used as a division value at the time of encoding) is searched, and the frequency conversion coefficient C ″ _j obtained by the search is multiplied by the division result C ′ _i (step T13).

After the multiplication process in step T13 or when it is determined in step T12 that the division result C ′ _i has been processed (step T12; YES), the counter value i is incremented (step T14), and the current counter value i Steps T11 to T13 are repeated. If it is determined in step T11 that all frequency conversion coefficients in one frame have been processed (step T11; YES), the coefficient multiplication process ends.

  As described above, according to the present embodiment, the division result is concentrated in the vicinity of ± 1 by dividing the frequency conversion coefficient for each unit in one frame by another frequency conversion coefficient most similar to each unit. Thus, the encoding efficiency can be improved.

  Further, by using the square error of the frequency conversion coefficient as an index when searching for the most similar frequency conversion coefficient, the search process can be performed by simple calculation.

<Modification>
In the above-described embodiment, the method of searching for the most similar frequency conversion coefficient by calculating the square error between the frequency conversion coefficients is shown. In this modification, a method of once performing division processing using each frequency conversion coefficient as a division value and searching for the most similar frequency conversion coefficient based on the bias (frequency distribution) of the division result value will be described.

Frequency conversion coefficients of interest and C _i, C _1, C 2 frequency conversion coefficients to be searched in the C _i and the same _frame, ..., and C _p. The C _i, C _1, ..., the result of dividing the respective _{C p, D i1, ...,} D ip, D ik = {d ikl | l = 1, ..., m} When, D _ik, d _ikl Is as shown in Equation (4).

When the calculated division result d _ikl as in Equation (4) is a value which appears most biased alpha, D _i1, ..., searching highest D _ik is number of the d _ikl = alpha in D _ip The corresponding C _k is the frequency conversion coefficient most similar to C _i . Even when this search method is used, the encoding efficiency can be improved.

  With reference to the flowchart of FIG. 6, the coefficient division processing in this modification will be described.

First, the frequency conversion coefficient C ₁ for the first (first in time) unit in the frame is divided by each of all other frequency conversion coefficients C _{2 to} C _n in the same frame ( Step S20). In step S20, a division result with i = 1 and k = 2 to n in the equation (4) is obtained.

Then, the frequency conversion coefficient frequency distribution is the most biased division result value in step S20 is retrieved as the most similar frequency transform coefficients C ₁ (step S21). The frequency conversion coefficient and coefficient number obtained by the search in step S21 are to be encoded in the encoding process of FIG.

  Next, the counter value i for designating the coefficient number of the frequency conversion coefficient is set to 1 (step S22). Hereinafter, an encoded frequency conversion coefficient obtained by searching for the most similar frequency conversion coefficient is set as a “processed” frequency conversion coefficient.

Next, it is determined whether or not all frequency conversion coefficients in one frame have been processed (step S23). If it is determined in step S23 that all the frequency conversion coefficients in one frame have not been processed (step S23; NO), it is determined whether or not the frequency conversion coefficient C _i of coefficient number _i has been processed. (Step S24).

If it is determined in step S24 that the frequency conversion coefficient C _i has not been processed (step S24; NO), C _i is divided by all the processed frequency conversion coefficients as shown in equation (4) ( Step S25). Then, based on the frequency distribution of the division result value in step S25, the frequency conversion coefficient frequency distribution is the most biased value of the division result is retrieved as the most similar frequency transform coefficients C _i (step S26). In the encoding process of FIG. 3, the result of dividing C _i by the most similar frequency transform coefficient and the coefficient number of the similar frequency transform coefficient (division value) are to be encoded.

After the search in step S26 or in step S24, if it is determined that the frequency conversion coefficient C _i has been processed (step S24; YES), the counter value i is incremented (step S27), and the current counter value i is determined. Steps S23 to S26 are repeated. If it is determined in step S23 that all frequency conversion coefficients within one frame have been processed (step S23; YES), the coefficient division process ends.

The block diagram (a) which shows the structure of the encoding apparatus which concerns on embodiment of this invention, and the block diagram (b) which shows the structure of a decoding apparatus. The flowchart which shows the coefficient division process performed in the encoding apparatus of FIG. The flowchart which shows the encoding process performed in the encoding apparatus of FIG. 3 is a flowchart showing a decoding process executed in the decoding device of FIG. 1. 3 is a flowchart showing coefficient multiplication processing executed in the decoding device of FIG. 1. The flowchart which shows the coefficient division process in the modification of this embodiment.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Frame conversion part 2 Frequency conversion part 3 Division part 4 Encoding part 5 Decoding part 6 Multiplication part 7 Frequency reverse conversion part 8 Frame synthetic | combination part 100 Encoding apparatus 200 Decoding apparatus

Claims (6)

Framing means for dividing the input signal into frames;
Coefficient calculation means for performing frequency conversion on the input signal to calculate a frequency conversion coefficient for each unit constituting one frame;
The most similar frequency transform coefficients to the first frequency conversion coefficients of one unit in one frame, a first retrieving means for retrieving from the other frequency conversion coefficients of one unit in the same frame,
First encoding means for encoding the most similar frequency conversion coefficient for one unit searched by the first search means and number information indicating the number of the frequency conversion coefficient for one unit in the frame. When,
A designation means for sequentially designating a frequency conversion coefficient for one unit to be processed in a frame;
Second search means for searching a frequency conversion coefficient most similar to the frequency conversion coefficient for one unit specified by the specifying means from among the already encoded frequency conversion coefficients for one unit;
Division means for dividing the frequency conversion coefficient for one unit to be processed by the most similar frequency conversion coefficient for one unit searched by the second search means;
Second encoding means for encoding the division result by the dividing means and the number information of the frequency conversion coefficient for one unit used for division by a predetermined encoding method;
Until the encoding of all the frequency conversion coefficients for one unit constituting one frame is completed, the processing of the specifying means, the second searching means, the dividing means, and the second encoding means is repeatedly executed. And a control means . An error calculating means for calculating an error between the predetermined one unit of frequency conversion coefficient and another one unit of frequency conversion coefficient in the same frame;
The encoding apparatus according to claim 1, wherein the first and second detection means detect the most similar frequency transform coefficient based on the error calculated by the error calculation means. A decoding device for decoding a signal encoded by the encoding device according to claim 1,
The signal encoded by the first encoding means of the encoding device is decoded, and the frequency conversion coefficient for one unit and the number information indicating the number of the frequency conversion coefficient for one unit in the frame Obtaining first decoding means;
Second decoding means for decoding the signal encoded by the second encoding means of the encoding device to obtain the division result and number information of the frequency conversion coefficient for one unit used for the division;
Multiplication means for multiplying the division result obtained by the second decoding means by a frequency transform coefficient used as a division value at the time of encoding specified by the number information ;
Inverse transform means for performing frequency inverse transform on the signal obtained by the multiplication means;
A decoding apparatus comprising:   A framing step for dividing the input signal into frames;
  A coefficient calculating step for performing frequency conversion on the input signal to calculate a frequency conversion coefficient for each unit constituting one frame;
  A first search step of searching for a frequency conversion coefficient most similar to the frequency conversion coefficient for the first unit in one frame from the frequency conversion coefficients for another one unit in the same frame;
  A first encoding step for encoding the most similar frequency conversion coefficient for one unit searched in the first search step and number information indicating the number of the frequency conversion coefficient for one unit in the frame. When,
  A designation step for sequentially designating a frequency conversion coefficient for one unit to be processed in a frame;
  A second search step for searching a frequency conversion coefficient most similar to the frequency conversion coefficient for one unit specified in the specifying step from among the already encoded frequency conversion coefficients for one unit;
  A division step of dividing the frequency conversion coefficient for one unit to be processed by the most similar frequency conversion coefficient for one unit searched by the second search step;
  And a second encoding step that encodes the division result of the division step and number information of the frequency conversion coefficient for one unit used for the division by a predetermined encoding method, and includes all 1s constituting one frame. An encoding method comprising repeatedly executing the processing of the specifying step, the second search step, the division step, and the second encoding step until the encoding of the frequency conversion coefficient for a unit is completed. . An error calculating step of calculating an error between the predetermined one unit of frequency conversion coefficient and another one unit of frequency conversion coefficient in the same frame;
5. The encoding method according to claim 4, wherein the first and second detection steps detect the most similar frequency transform coefficient based on an error calculated by the error calculation unit . A decoding method for decoding a signal encoded by the encoding method according to claim 4, comprising:
  The signal encoded in the first encoding step of the encoding method is decoded, and the frequency conversion coefficient for one unit and number information indicating the number of the frequency conversion coefficient for one unit in the frame are obtained. Obtaining a first decoding step;
  A second decoding step of decoding the signal encoded in the second encoding step of the encoding method to obtain a division result and frequency conversion coefficient number information for one unit used for the division;
  A multiplication step of multiplying the division result obtained by the second decoding step by a frequency transform coefficient used as a division value at the time of encoding specified by the number information;
  A decoding method comprising: an inverse transform step for performing frequency inverse transform on the signal obtained by the multiplication step.

Images