JP2007219438A - Code book selection method and code book selection device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To select an optimum code book used for codes. <P>SOLUTION: The number of coded bits is stored in a bit number table specified by a code book number and a scale factor band number; a bit cost value is stored in accumulative value memory as an accumulative value; a value adding the bit cost value to the minimum accumulative value is stored in threshold value memory. When the accumulative value is larger than the threshold value, the accumulative value is stored in the accumulative value memory as a new accumulative value, and also a code book number for this minimum accumulative value is stored in a link table. Meanwhile, when the accumulative value is the threshold value or smaller, the last code book number is stored in the link table as a code book number for this time; and a value adding the accumulative value to the number of bits is updated as a new accumulative value. The optimum code book number is determined by referring to the link table and sequentially pursuing the code book number, with which the number of bits in the bit number table corresponds to the minimum number of bits, from a high order scale factor band number. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to a code book selection method and a code book selection apparatus, and is used for, for example, selection of a code book of AAC, for example, and a code book selection method and code book selection for searching for the minimum number of bits for expressing a quantized spectrum. Relates to the device.

  To date, a great number of audio signal encoding methods have been developed. Particularly recently, ISO 13818-7, which has been internationally standardized by ISO / IEC, has been recognized and evaluated as an encoding method with high sound quality and high efficiency. This encoding method is called AAC.

  In recent years, AAC has also been adopted in the MPEG-4 standard, and the MPEG4-AAC system having some extended functions is defined for ISO 13818-7.

  Here, Patent Document 1 has been reported as a method of combining (merging) scale factor bands into sections. Hereinafter, Patent Document 1 will be described.

  The invention described in Patent Document 1 is a method for merging a scale factor band (hereinafter referred to as SFB) of an audio signal into a plurality of groups each called a section, and for encoding each SFB. Find all possible Huffman codebooks of, calculate the total bit cost required to represent the spectral coefficients of the frame using the Huffman codebook with the lowest bit cost (minimum code book), and the same Search for subsequent SFBs with the minimum codebook, merge them as one section, set the reference total bit cost equal to the total bitcost, and the section to be merged is codebook 0 or intensity stereo coded codebook Check whether to use. Then, over all pairs of subsequent sections, calculate the new total bit cost obtained by temporarily merging only one pair, of which the pair of sections that yields the lowest new total bit cost is found and this minimum If the new total bit cost is less than the total bit cost, reset the total bit cost equal to the new total bit cost. On the other hand, if the total bit cost is lower than the reference total bit cost (bit saving), the pair of sections is merged permanently, and the new section obtained by merging the pair of sections and the adjacent section have the same minimum code. Check if you have a book. If the new section and the adjacent section have the same minimum codebook, they are merged to form a new section. Set the total bit cost as the new reference total bit cost and repeat until no bit savings are provided.

As described above, Patent Document 1 describes a method of combining (merging) SFBs into sections, and in particular, can be expressed using the same code book.
JP 2001-188563 A

  By the way, in the process of Huffman encoding the quantized spectrum, the total number of bits changes depending on the selection of the code book. That is, if the optimum code book is selected, the total number of bits is minimized, but if an inappropriate code book is selected, the total number of bits increases.

  If the code book determination method simply selects a code book that gives the minimum bit for each SFB, the number of sections increases, resulting in a bit cost. The waste of bits is reduced by the quantization width of the spectrum. As a result, the quantization noise is relatively increased and the sound quality is deteriorated.

  Therefore, the code book must be selected so that the number of bits is minimized as a whole in consideration of the bit cost.

  In contrast, the technique described in Patent Document 1 is configured to sequentially reduce the total number of bits used for encoding using the same code book, but the minimum number (logically) There is a problem that the sound quality is not necessarily reduced.

  Therefore, it is desired to provide a method for selecting an optimum code book in order to minimize the total number of bits, suppress a relative increase in quantization noise, and prevent deterioration in sound quality.

  The present invention has been made in view of the above, and an object thereof is to provide a code book selection method and a code book selection apparatus capable of selecting an optimum code book used for encoding.

  In order to solve the above problem, the invention according to claim 1 is a bit in which the number of bits resulting from encoding for all code books for each scale factor band is designated by the code book number and the scale factor band number. A step of storing in the number table and storing the bit cost value as an accumulated value in the accumulated value memory for initialization, and a value obtained by adding the bit cost value to the minimum value of the accumulated value from the accumulated value memory In the threshold memory as a threshold value, and when the accumulated value from the accumulated value memory is larger than the threshold value from the threshold memory, the accumulated value is stored in the accumulated value memory as a new accumulated value; The code book number for this minimum accumulated value is stored in the link table. On the other hand, if the accumulated value is less than or equal to the threshold value, the previous code book number from the link table is stored as the current code book number. Storing in the link table as a signal, updating a value obtained by adding the number of bits from the bit number table and the accumulated value from the accumulated value memory as a new accumulated value, and the link table And referring to the code number corresponding to the minimum number of bits stored in the bit number table in order from the higher scale factor band number to determine the optimum code number. And

  In order to solve the above problem, the invention according to claim 2 is a bit specified by the code number and the scale factor band number for the number of bits encoded for all code books for each scale factor band. Means for storing and initializing a bit cost value as an accumulated value in an accumulated value memory, and a value obtained by adding the bit cost value to the minimum accumulated value from the accumulated value memory; Is stored in the threshold value memory as a threshold value, and when the accumulated value from the accumulated value memory is larger than the threshold value from the threshold value memory, the accumulated value value is stored in the accumulated value memory as a new accumulated value; The code number for the minimum accumulated value is stored in the link table. On the other hand, if the accumulated value is less than or equal to the threshold value, the previous code number from the link table is set as the current code number. Means for storing in the link table, means for updating the value obtained by adding the bit number from the bit number table and the accumulated value from the accumulated value memory as a new accumulated value, and referring to the link table. And a means for determining an optimum code number by tracking the code number corresponding to the minimum number of bits in the bit number table in order from the higher scale factor band number. .

  In order to solve the above-mentioned problem, the invention described in claim 3 inputs the number of bits indicating the result of encoding for all the code books for each scale factor band, and enters the code number and the scale factor band number. Input storage means for storing in the bit number table specified by the above, initialization means for storing and initializing the bit cost value in the accumulated value memory as an accumulated value corresponding to each code number, and the accumulated value Threshold value calculation means for adding the bit cost value to the minimum value of the accumulated value read from the memory and storing the addition result value as a threshold value in the threshold value memory; the accumulated value read from the accumulated value memory and the threshold value Compared with the threshold value read from the memory, if this accumulated value is larger than the threshold value, it is stored in the accumulated value memory as a new accumulated value, and the minimum accumulated value is The code number is stored in the link table. On the other hand, if the accumulated value is less than or equal to the threshold value, the code number specified by the previous scale factor band number stored in the link table is read, and this code number Is stored in the link table as the current code book number, and the bit number corresponding to each code book number in the scale factor band number is read from the bit number table and accumulated from the accumulated value memory. Accumulated value update means for reading the calculated value, adding the number of bits corresponding to each code number and the accumulated value, and updating the added value as a new accumulated value corresponding to each code number, and the link Until the table is completed, the threshold value calculation means, the accumulated value correction link determination means, and the accumulated value update means are sequentially repeated for each scale factor band number. Referring to the completed link table and the repetition means to be executed, the code number corresponding to the minimum number of bits stored in the bit number table is traced in order from the higher scale factor band number in order and optimized. And a code number determination means for selecting and determining a code number.

  According to the present invention, it is possible to select an optimum code book to be used for encoding, so that the number of bits in encoding can be saved, and quantization quality can be reduced to improve sound quality.

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

  FIG. 1 is a block diagram illustrating a configuration of an encoding device 11 according to an embodiment applicable to the code book selection device according to the present invention, and FIG. 2 illustrates a configuration of a quantization unit 15 of the encoding device 11. It is a block diagram.

  In FIG. 1, the time-frequency conversion unit 13, the quantization unit 15, and the bit stream generation unit 17 constituting the encoding device 11 may be implemented by software processing or program processing in a DSP (Digital Signal Processor). May be executed.

  In FIG. 1, s is a time series signal s of an audio signal to be encoded. The time-frequency converter 13 performs a modified discrete cosine transform (MDCT) on the input time series signal s and outputs a spectrum f. The quantization unit 15 quantizes the spectrum f output from the time-frequency conversion unit 13 and outputs a quantized spectrum q. The bit stream generation unit 17 generates and outputs the bit stream bs according to the syntax determined from the quantized spectrum q output from the quantization unit 15.

  In addition, like MPEG-2 AAC and MPEG-4 AAC (hereinafter, these two AACs are collectively referred to as AAC), a PNS (Perceptual Noise Shaping) is provided between the time-frequency conversion unit 13 and the quantization unit 15. ) A spectral processing unit such as an encoding tool or an IS (Intensity Stereo) encoding tool may be inserted.

  In audio coding using human auditory characteristics in the quantizing unit 15, a method of dividing the quantized spectrum into several bands may be used. There is also a process for entropy encoding the quantized spectrum for each band.

  For example, in AAC, 1024 quantized spectra are divided into 40 to 50 bands according to the sampling frequency, and Huffman coding is performed on the quantized spectra belonging to each band. (However, the above numerical example is for a long window. In the case of a short window, 128 quantized spectra are divided into 10 to 15 bands.) This band is called a scale factor band (hereinafter referred to as SFB). There are a plurality of Huffman codebooks (codebooks), and there are codebooks that can be used and codebooks that cannot be used depending on the constraint conditions. For each SFB, it is possible to select which one to use.

  The number of the selected code book is encoded separately and described in the AAC stream. At this time, if the same code number is selected between adjacent SFBs, they can be described together, and this is called section data, and the process of grouping is called sectioning. The section data is composed of a code book number and an SFB range in which the code book is selected, and is described in order from the low frequency side. In order to describe one section data, 9 [bit] is required in the case of a long window, and 7 [bit] is required in the case of a short window. This is hereinafter referred to as a bit cost.

  Next, a block diagram of the quantization unit 15 will be described with reference to FIG.

  The quantizer 21 divides the given spectrum f by the corresponding scale factor sf and outputs a quantized spectrum q ′. The quantized spectrum q ′ may be required to be re-quantized by bit control, and is not necessarily coincident with the quantized spectrum q that is finally output from the quantizing unit 15 and is therefore denoted as q ′.

  The Huffman code book candidate determination unit 23 extracts and determines code book candidates to be subjected to Huffman coding for each SFB, and outputs an output c describing which code book should be tried for each SFB. Output. Since the Huffman code book is restricted in use by the maximum absolute value in the SFB, the code book candidate needs to be determined by the Huffman code book candidate determination unit 23. The Huffman code book candidate determination unit 23 also plays a role of reducing code book candidates for the purpose of reducing the processing load.

  The Huffman coding unit 27 performs Huffman coding on all code books described in c for each SFB, and outputs the resulting number of bits bt. The number of bits bt is a table, and describes how many bits the quantization spectrum in a certain SFB becomes when Huffman coding is performed with a certain code book.

  The code book determination unit 29 determines which code book number to select for each SFB based on the value described in the table of the number of bits bt output from the Huffman coding unit 27. The determined code book hcb and the total bit number b of the encoding result are output. The total number of bits b includes the above-mentioned bit cost due to code book switching.

  The determination unit 31 determines the range of the total number of bits b given from the code book determination unit 29. If the total number of bits b is within the predetermined allowable range, the determination unit 31 switches the permission signal ji. 33, the switch 33 is turned on, and the quantized spectrum q ′ output from the quantizer 21 is output as the quantized spectrum q to the bitstream generation unit 17 via the switch 33. At the same time, the code book hcb is separately described in the stream as sub information. However, when the total number of bits b given from the code book determination unit 29 is outside this allowable range, the determination unit 31 gives the non-permission signal jn to the scale factor updater 35 and performs requantization.

  The scale factor updater 35 increases / decreases the scale factor sf used for the previous quantization, and then gives the updated scale factor sf to the quantizer 21 for requantization.

  Next, the operation of the quantization unit 15 will be described with reference to FIG.

  First, when the spectrum f is output from the time-frequency conversion unit 13 to the quantizer 21, the quantizer 21 divides the given spectrum f by the corresponding scale factor sf and outputs a quantized spectrum q ′. To do. Next, the Huffman code book candidate determination unit 23 extracts and determines code book candidates to be subjected to Huffman coding for each SFB, and outputs which code book should be tried for each SFB. c is output. Next, the Huffman coding unit 27 performs Huffman coding on all code books described in c for each SFB, and outputs the resulting number of bits bt. Next, the code book determination unit 29 determines which code book number is selected for each SFB based on the value described in the table of the number of bits bt output from the Huffman coding unit 27. The determined code book hcb and the total bit number b of the encoding result are output.

  Here, the determination unit 31 determines the range of the total number of bits b given from the code book determination unit 29, and when the total number of bits b is within a predetermined allowable range, jy is given to the switch 33, the switch 33 is turned on, and the quantized spectrum q ′ output from the quantizer 21 is output to the bitstream generation unit 17 via the switch 33 as the quantized spectrum q. At the same time, the code book hcb is separately described in the stream as sub information.

  On the other hand, when the total number of bits b given from the code book determination unit 29 is outside the allowable range, the determination unit 31 gives the non-permission signal jn to the scale factor updater 35 and performs requantization. Next, the scale factor updater 35 increases / decreases the scale factor sf used for the previous quantization and then gives the updated scale factor sf to the quantizer 21 to perform requantization.

  Next, the memory configuration used for the processing in the code book determination unit 29 will be described with reference to the explanatory diagram shown in FIG. In the following description, the number of code books is represented as n, and the number of SFBs as m. As shown in the explanatory diagram of FIG. 4, the bit number table 41, the accumulated value memory 43, the threshold value memory 45, and the link table 47 are required to implement the processing in the code book determination unit 29.

  The bit number table 41 is a memory that stores the Huffman-encoded bit number bt in correspondence with the code number t and the scale factor band number i, and receives the bit number bt indicating the Huffman-encoded result. The code is stored in the position specified by the code number t and the scale factor band number i.

  The accumulated value memory 43 is a memory having a storage capacity as large as the number of code books, and is represented by a (1) to a (n). a (t) [where t is an integer from 1 to n] represents the number of bits when it is assumed that the code book number t is selected.

  The link table 47 is a two-dimensional memory for storing code number, and requires a maximum storage capacity of (number of code lines) × (number of SFBs). 1,1) to k (m, n). k (i, t) [where i is an integer from 1 to m and t is an integer from 1 to n] is assumed when the code number t is selected in the i-th SFB In the previous i-1 th SFB, it is shown which code book is optimal (bits are minimized).

  Next, the operation of the code book determination unit 29 shown in FIG. 2 will be described with reference to the flowchart shown in FIG. 3 and the explanatory diagrams shown in FIGS.

  First, in step S5, the code book determination unit 29 inputs the bit number bt indicating the result of Huffman coding for all the code books for each scale factor band by the Huffman coding unit 27 and inputs the code book number. It is stored in the bit number table 41 specified by t and the scale factor band number i.

  Next, in step S10, the code book determination unit 29 initializes the accumulated values a (1) to a (n) with a bit cost. That is, if the frame is a long window, “9” is assigned to the accumulated values a (1) to a (n). Hereinafter, with respect to a certain i (i is an integer of 1 to m) th SFB, the processes shown in steps S20 to S50 are repeated.

Next, in step S20, the code book determination unit 29 obtains the minimum value among the accumulated values a (1) to a (n), adds the bit cost to the obtained minimum accumulated position, and obtains this. The threshold value is thr. That is, the threshold thr in the case of a long window is
thr = min {a (1),. . . , A (n)} + 9 (Equation 1)
The threshold thr in the case of a short window is
thr = min {a (1),. . . , A (n)} + 7 (Equation 2)
It is. The code number t for the minimum accumulated value is tmin. Note that when there are a plurality of minimum values tmin, it is not the essence of the present invention which one is selected.

Next, in step S30, the code book determination unit 29 compares the accumulated values a (1) to a (n) with the threshold value thr in magnitude. here,
(1) If a (t)> thr,
It is corrected as a (t) = thr and stored in the link table 47 as k (i, t) = tmin. That is, when the accumulated value a (t) is larger than the threshold value thr, the threshold value thr is substituted into the accumulated value memory 43 and stored as a new accumulated value. At the same time, the code book number tmin with respect to the minimum accumulated value a (t) is stored in the link table 47 as the link k (i, t) = tmin formed by the SFB number i and the code book number t.

(2) When a (t) ≦ thr,
The code number designated by the previous scale factor band number (i-1) stored in the link table 47 is read without correcting the accumulated value, and the code number t is read as the current code number k (i, t) and stored in the link table 47.

  Here, the meanings of the processing divided into cases (1) and (2) are as follows.

  Assuming that the code number t is used in the i-th SFB, in the case of (1): If it is determined that a (t)> thr, the code number tmin is set in the immediately preceding (i-1) -th SFB. This indicates that the number of bits is smaller when the code number t is selected in the i-th SFB even if the bit cost is considered.

  On the other hand, in the case of (2): if it is determined that a (t) ≦ thr, it is more costly to use the code number t without changing the code number from the immediately preceding (i−1) th SFB. Not only does it show that the overall number of bits can be saved.

  Therefore, when i = 0, the link table n value has no meaning (because there is no immediately preceding SFB) and is merely convenient.

  Next, in step S40, the code book determining unit 29 adds the bit number bt (i, t) to the accumulated value a (t). The number of bits bt is the output of the Huffman encoder 27 shown in FIG. 2, and the number of bits when the i-th SFB is encoded with the code book t is given by bt (i, t).

  It is assumed that “∞” is assigned for the sake of convenience for code books that cannot be selected by the Huffman code book candidate determination unit 23. Use conditional branches such as if statements.

  Next, in step S50, the code book determination unit 29 determines whether the processing has been completed for all SFBs. That is, if the i-th SFB is the last SFB to be encoded, that is, if i = m, the process proceeds to step S60 in order to perform end processing.

  Otherwise, the process proceeds to step S55, and the code book determination unit 29 updates “i” to “i + 1” and returns to step S20 to execute the process for the next i + 1-th SFB. That is, until the link table 47 is completed, Steps S20 to S40 are repeatedly executed sequentially for each scale factor band number i.

Finally, in step S60, the code book determination unit 29 obtains the minimum value from the accumulated values a (1) to a (n) stored in the accumulated value memory 43, and this becomes the minimum bit b of the output. . In addition, the code number which becomes the minimum bit b is tmin, that is, a (tmin) = b. At this time, the code book number for the m-th terminal SFB is determined to be tmin, and the code book number corresponding to the SFB numbers from 1 to (m−1) which is the minimum number of bits is stored in the link table 47 from the higher SFB number. On the contrary, it is determined sequentially by tracing. If you write this process as a formula,
hcb (m) = tmin (Equation 3)
hcb (i) = k (i + 1, hcb (i + 1)) (Equation 4)
It becomes. Here, hcb (i) is the code number of the i-th SFB.

  That is, referring to the completed link table 47, the code number corresponding to the minimum number of bits stored in the bit number table 41 is traced in order from the higher SFB number to select the optimum code number. To decide.

  Next, processing in each step of the flowchart shown in FIG. 3 will be described with reference to specific examples.

  The code book is assumed to be 12 books from “0” to “11”, and the SFB is assumed to be 12 bands from “0” to “11”.

  First, in step S5, the code book determination unit 29 inputs the bit number bt indicating the result of Huffman coding for all the code books for each scale factor band by the Huffman coding unit 27 and inputs the code book number. It is stored in the bit number table 41 specified by t and the scale factor band number i.

  The number of bits resulting from each Huffman coding is given as shown in the table “number of bits” shown in FIG. In this table, the symbol “∞” indicates a code number that cannot be selected by the Huffman code book candidate determination unit 23. In practice, a sufficiently large value is entered or a conditional branch such as an “if” statement is executed. Use.

  Next, in step S10, the accumulated values a (1) to a (n) are initialized. Assuming that the frame is a short window, the bit cost is “7”, and as shown in FIG. 5, the bit cost “7” is an accumulated value with each of accumulated values a (1) to a (n). It is stored in the memory 43 and initialized.

  Next, in step S20, the bit cost “7” is added to the minimum value of the accumulated values a (1) to a (n) to obtain a threshold value. In this case, as shown in FIG. 6, since the accumulated values a (1) to a (n) are all the same value “7”, tmin = 0 is set from the code number where the accumulated value is the minimum value. The threshold is set to “14” (= 7 + 7).

  Next, in step S30, the magnitudes of the accumulated values a (1) to a (n) and the threshold value thr are compared. As shown in FIG. 7, in this case, since there is no accumulated value exceeding the threshold value thr = 14, the accumulated value is not corrected. Further, the link of the code number t = 0 to 11 for the SFB number i = 0 is stored in the link table 47 as k (i, t) = t.

  Next, in step S40, the number of bits of the SFB shown in FIG. 8 (A in the figure) is added to the accumulated value shown in FIG. 7, and the addition result is stored in the accumulated value memory 43 as an accumulated value. Update.

  Next, in step S50, since the processing related to all the SFBs has not been completed, the process proceeds to step S55, the process is moved to the next “i + 1” -th SFB, and the above-described processes are repeated.

  Here, in step S20, as shown in FIG. 9, the bit cost “7” is added to the minimum value “23” of the accumulated value to obtain the threshold value “30”. Tmin = 10 is selected from the code number where the accumulated value becomes the minimum value, and the threshold value is set to “30” and stored in the threshold value memory 45.

  Next, in step S30, the accumulated values a (1) to a (7) corresponding to the code number t = 0 to 6 are “∞” and exceed the threshold value thr = “30”. As shown, the accumulated values a (1) to a (7) are corrected so that the link source is tmin = 10. Otherwise, the accumulated value is not corrected and stored in the link table 47 as k (i, t) = t from itself.

  Next, in step S40, as shown in FIG. 11, the number of bits of the SFB (B in the figure) is added to the accumulated value, and this added value is stored as an accumulated value in the accumulated value memory 43 and updated. To do.

  Next, in step S50, since the processing related to all the SFBs has not been completed, the process proceeds to step S55, the process is moved to the next “i + 1” -th SFB, and the above-described processes are repeated.

  Similarly, in step S20, as shown in FIG. 12, the bit cost “7” is added to the minimum value “36” of the accumulated value to obtain the threshold value “43”. Tmin = 10 is selected from the code number with the minimum accumulated value, and the threshold value is set to “43” and stored in the threshold value memory 45.

  Next, in step S30, as shown in FIG. 13, the accumulated values a (1) to a (7) and a (9) corresponding to the code numbers 0 to 7 and 9 exceed the threshold “43”. Therefore, they are corrected and tmin is stored in the link table 47. Otherwise, no correction is made, and k (i, t) = t from itself is stored in the link table 47.

  Next, in step S40, as shown in FIG. 14, the number of bits of the SFB (C in the figure) is added to the accumulated value, and the addition result is stored in the accumulated value memory 43 as an accumulated value and updated. To do.

  Next, in step S50, it is determined whether all SFBs have been completed. Since the processes related to all SFBs have not been completed, the process proceeds to step S55, and the process moves to the next “i + 1” -th SFB. Thereafter, the processes of steps S20 to S40 described above are similarly repeated.

  If it is determined in step S50 that all SFBs have been completed, the process proceeds to step S60. In step S60, at this time, when the SFB is accumulated up to the terminal (i = 11) SFB, the minimum value of the accumulated value is obtained as shown in FIG. In this case, tmin = 0 and the code number at the end is determined to be 0.

The other code number hcb (i) is in accordance with Equation 5,
hcb (i) = k (i + 1, hcb (i + 1)) (Equation 5)
It becomes.

  The code book number corresponding to the SFB numbers from 1 to (m−1), which is the minimum number of bits, is obtained by tracing the link table 47 backward from the upper SFB number. That is, the number i is sequentially substituted from i = 11 to i = 0, and the code book number is selected from the completed link table 47 shown in FIG. That is, referring to the completed link table 47, the code number corresponding to the minimum number of bits stored in the bit number table 41 is traced in order from the higher SFB number to select the optimum code number. To decide.

  As a result, as shown in FIG. 15, selection of the optimum code number is “8”, “8”, “8”, “8”, “8” in order from the low frequency side to the high frequency side. , “8”, “6”, “6”, “6”, “6”, “0”, “0”, and the number of bits after encoding is 218 [bits]. Here, for example, the first “8 8 8 8 8 8” indicates session 1, the next “6 6 6 6” indicates session 2, and the last “0 0” indicates session 3.

  If a Huffman code book is selected according to the method of the present invention, a given spectrum can be expressed with a minimum number of bits. If bits can be saved by Huffman coding of the spectrum, the quantization width can be made narrower, and sound quality improvement such as reduction of quantization noise can be realized.

  In the embodiment of the present invention, a case where an AAC PNS encoding tool or an IS encoding tool is used is not described, but a hypothesis indicating that a PNS encoding tool or an IS encoding tool is used is used. The same can be applied by adding a typical code number (13 to 15).

  That is, in the present invention, it is assumed that the number of bits after Huffman coding is given to all codebooks that can be selected, and the present invention provides the optimum codebook number from the table indicating the number of bits. It is characterized by selecting.

  In the present embodiment, the audio signal is encoded by Huffman encoding. However, the present invention is not limited to Huffman encoding, and can be applied to other encoding methods.

It is a block diagram which shows the structure of the encoding apparatus 11 used as one Embodiment applicable to the code book selection apparatus based on this invention. 3 is a block diagram illustrating a configuration of a quantization unit 15 of the encoding device 11. FIG. 6 is a diagram for explaining an operation of a code book determination unit 29. FIG. 5 is a diagram for explaining a configuration of a memory used for processing in a code book determination unit 29. FIG. It is a figure for demonstrating the operation | movement in step S10. It is a figure for demonstrating the operation | movement in step S20. It is a figure for demonstrating the operation | movement in step S30. It is a figure for demonstrating the operation | movement in step S40. It is a figure for demonstrating the operation | movement in step S20. It is a figure for demonstrating the operation | movement in step S30. It is a figure for demonstrating the operation | movement in step S40. It is a figure for demonstrating the operation | movement in step S20. It is a figure for demonstrating the operation | movement in step S30. It is a figure for demonstrating the operation | movement in step S40. It is a figure for demonstrating the operation | movement in step S60.

Explanation of symbols

DESCRIPTION OF SYMBOLS 11 ... Coding apparatus, 11a ... CPU, 11b ... Memory 11b, 13 ... Time-frequency conversion part, 15 ... Quantization part, 17 ... Bit stream generation part, 21 ... Quantizer, 23 ... Huffman code book candidate determination part , 27 ... Huffman encoding unit, 29 ... Code book determination unit, 31 ... Determination unit, 33 ... Switch, 35 ... Scale factor updater, 41 ... Bit number table, 43 ... Accumulated value memory, 45 ... Threshold memory, 47 ... Link table

Claims (3)

The number of bits resulting from encoding for all code books for each scale factor band is stored in the bit number table specified by the code book number and the scale factor band number, and the bit cost value is accumulated as an accumulated value. Storing and initializing in arithmetic memory;
Storing the value obtained by adding the bit cost value to the minimum value of the accumulated value from the accumulated value memory as a threshold value in the threshold memory;
If the accumulated value from the accumulated value memory is larger than the threshold value from the threshold memory, the accumulated value is stored in the accumulated value memory as a new accumulated value, and the code number for this minimum accumulated value is also stored. Storing in the link table, on the other hand, if the accumulated value is less than or equal to the threshold, storing the previous code book number from the link table as the current code book number in the link table;
Updating a value obtained by adding the bit number from the bit number table and the accumulated value from the accumulated value memory as a new accumulated value;
Referring to the link table, and determining the optimum code number by tracking the code number corresponding to the minimum number of bits stored in the bit number table in order from the higher scale factor band number; A code book selection method comprising: The number of bits resulting from encoding for all code books for each scale factor band is stored in the bit number table specified by the code book number and the scale factor band number, and the bit cost value is accumulated as an accumulated value. Means for storing and initializing in arithmetic memory;
Means for storing in the threshold memory a value obtained by adding the bit cost value to the minimum value of the accumulated value from the accumulated value memory;
If the accumulated value from the accumulated value memory is larger than the threshold value from the threshold memory, the accumulated value is stored in the accumulated value memory as a new accumulated value, and the code number for this minimum accumulated value is also stored. Means for storing in the link table, on the other hand, if the accumulated value is less than or equal to the threshold, storing the previous code book number from the link table as the current code book number in the link table;
Means for updating a value obtained by adding the bit number from the bit number table and the accumulated value from the accumulated value memory as a new accumulated value;
Means for referring to the link table and determining the optimum code number by tracking the code number corresponding to the minimum number of bits stored in the bit number table in order from the higher scale factor band number; A code book selection device comprising: Input storage means for inputting the number of bits indicating the result of encoding for all the code books for each scale factor band and storing it in a bit number table designated by the code book number and the scale factor band number;
Initialization means for storing and initializing the bit cost value in the accumulated value memory as an accumulated value corresponding to each code number;
A threshold value calculating means for adding the bit cost value to the minimum value of the accumulated value read from the accumulated value memory and storing the addition result value as a threshold value in the threshold memory;
The accumulated value read from the accumulated value memory is compared with the threshold value read from the threshold memory, and when the accumulated value is larger than the threshold value, the accumulated value is set as a new accumulated value. The code number for the minimum accumulated value is stored in the link table as well as stored in the memory. On the other hand, if the accumulated value is less than or equal to the threshold value, the previous scale factor band number stored in the link table Accumulated value correction means for reading the designated code book number and storing this code book number as the current code book number in the link table;
Read the number of bits corresponding to each code number in the scale factor band number from the bit number table, read the accumulated value from the accumulated value memory, and add the number of bits and the accumulated value corresponding to each code number An accumulated value updating means for updating the added value as a new accumulated value corresponding to each code number;
Repeating means for repeatedly executing the threshold value calculating means, the accumulated value correction link determining means, and the accumulated value updating means sequentially for each scale factor band number until the link table is completed,
By referring to the completed link table, the code number corresponding to the minimum number of bits stored in the bit number table is traced in order from the higher scale factor band number, and the optimum code number is selected. A code book number selecting means for determining the code book number.

Images