JP2002057585A - Decoder - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a decoder which enables decrease in the retrieval frequency for decoding a Huffman code, can fast decode the Huffman code and also enables reduction in the table capacity. SOLUTION: This decoder is provided with a code retrieval table 10 where each record is composed of a decoding index, the code length and the number of samples, these records are divided into groups for each code length and the records included these groups are aligned in order of larger value of the corresponding Huffman code words, an input signal organization means 13 which inputs successively the Huffman coded signals and composes these signals into a retrieval bit string corresponding to the code length of the group to be retrieved, a code word generation means 14 which generates the largest Huffman code word in the group to be retrieved from the smallest Huffman code word and the number of samples included in the group and then generates the smallest Huffman code word of the next group from the difference value between the code length and that of the next group to be retrieved and the above largest Huffman code word, a comparison means 15 which compares the generated largest Huffman code word in the group with the composed input signals and a position calculation means 16 which calculates the entry position of the decoding index included in the group from the difference value obtained by the comparison.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a decoding device for decoding variable-length decoded data, and more particularly to a decoding device capable of decoding Huffman codes at high speed and reducing the capacity of a table. .

[0002]

2. Description of the Related Art Digitized data such as audio data and video data is compressed and then transmitted or recorded by encoding the contents according to a predetermined procedure. One of the techniques for reversibly compressing the data of such an information source is a variable-length coding method, and a representative coding method among them is known as a Huffman coding method. The Huffman code allocates an optimal code when the statistical properties of an information source are known. The Huffman code is a method of encoding data according to a symbol appearance probability distribution, and a short code is used for information having a high appearance probability. And assigning a long code to information with a low appearance probability to reduce the average bit length of the entire information source.

FIG. 5 is an explanatory diagram showing an example of Huffman coding. As shown in FIG. 9A, the appearance probability distribution of 16 types of indexes “0” to “15” of the information source is given in advance. For example, the appearance probability of the index “9” is “0”. .25 ", index" 1
"0" is "0.126", and index "8" is "0.126".
25 ",....

When a Huffman code is assigned to each of these indices, for example, a Huffman table as shown in FIG. In this example of the Huffman table, codes are assigned so as to give “0” to the shorter code length. As a result, the code word with the shortest code length (length) is assigned to the index “9” having the highest occurrence probability. (Codeword) “00” is assigned, and the shortest codewords “010” to “100” that can distinguish the index “9” from the indexes “10”, “8”, and “13” having the next highest occurrence probabilities Is assigned. Furthermore,
The indexes “0”, “1”, and “7” having the next highest appearance probabilities are the shortest code words “1010” to “11” that can distinguish the indexes “9”, “10”, “8”, and “13”.
00 ”is assigned. Similarly, a shorter codeword is assigned to an index with a higher appearance probability, and a longer codeword is assigned to an index with a lower appearance probability.

In order to decode such an encoded data signal, as shown in FIG. 5 (b), the data signal is converted using a Huffman table composed of a pair of a code word and an index. An index uniquely determined from the Huffman code word is searched from this table to obtain an original signal. Note that this Huffman table is sorted in ascending order of code length in advance, and further within the same code length is arranged in descending order of appearance probability in order to improve search efficiency.

The search procedure will be specifically described. Since the head code length of the Huffman table is 2 bits, when a coded signal is input, the head 2 bits are read and the read data is encoded. Compare with the word "00". If they match, the corresponding index “9” is extracted, and if they do not match, the code length of the next code is checked. FIG.
In the example of (b), since the code length of the next code is 3 bits, one more bit is read following the previously read 2 bits, and the 3-bit data is converted to the 3-bit code word “100”. "011" and "010". If it matches any one of them, the corresponding index is extracted, and if not, the next code length is further checked and the search is advanced.

As described above, while the number of bits to be searched is increased, the search is ultimately performed up to the maximum code length in the Huffman table, and if it exceeds this, a code error occurs.

In order to facilitate a search for each code length,
A search target flag may be set in the last data of a code group having the same code length.

As described above, in the conventional Huffman decoding, a Huffman table to be used is constituted by an array having higher-order information, and the Huffman table is sequentially compared with a code having a high appearance frequency.
Repeat the comparison process until they match. When the code word matches, the decoded signal value is obtained from the index corresponding to the code word.

[0010]

As described above, conventionally,
When decoding a Huffman code, it is necessary to sequentially search the table from the beginning for each code to be decoded using a Huffman table expanded on a memory, and thus the number of searches tends to increase ( The maximum number of searches is several times the number of entries in the table), and improvements in processing speed have been required. In particular, a playback device for audio or video, for example, decodes and plays in real time, so that a further improvement in processing speed is required.

The Huffman table used for searching also requires a large amount of information such as a code word, an index, a code length, and a search end flag, which results in an increase in table capacity. Reduction is required. In particular, in recent years, AV devices have been reduced in size and weight, and improvement of the memory capacity scale has become an important issue.

The present invention has been made in view of such a problem, and an object of the present invention is to reduce the number of searches for decoding a Huffman code and to decode a Huffman code at high speed. Another object of the present invention is to provide a decoding device capable of reducing the capacity of a table.

[0013]

According to a first aspect of the present invention, there is provided a decoding apparatus for decoding an input Huffman coded signal, comprising: a decoding index, a code length, a number of samples, , The records are grouped by the code length, and the records in the group are arranged in ascending order of the value of the corresponding Huffman code word, and the minimum Huffman code word of the first group is given as an initial value. Code search table, the input signal organizing means for sequentially inputting the Huffman coded signal, and organizing into a search bit string corresponding to the code length of the group to be searched, the minimum Huffman code word in the group to be searched and the number of samples. Generates the maximum Huffman code word in the group from the code length of the group and the difference value between the code length of the next search group and the A codeword generating means for generating a minimum Huffman codeword of the next group from the large Huffman codeword; a comparing means for comparing the generated input signal with the generated maximum Huffman codeword in the group; Position calculating means for calculating the entry position of the decoding index in the group from the difference value.

According to a second aspect of the present invention, the codeword generating means adds 1 to the maximum Huffman codeword,
Furthermore, the minimum Huffman code word of the next group is generated by shifting it to the right by the number of bits according to the difference value.

According to the decoding apparatus of the present invention, the code length, the number of samples, and the minimum value within the group, which is the initial value, are utilized by making use of the fact that the code values within the group of the same length are serialized when the Huffman table is created. Huffman codewords can be generated from codewords and the like. Compared to the Huffman codeword required in the conventional Huffman table, the number of samples required in the present invention can shorten the field length, resulting in a reduction in table capacity and memory saving. can do.

Further, it is determined whether or not the search target bit string exists in the group currently being searched using the generated Huffman code word, the number of samples, and the like, and if so, which entry in the group is located. It is calculated, and if it does not exist, the next group is searched. Therefore, the maximum number of searches can be set to (the number of groups + 1) times, and the processing time required for decoding can be reduced as compared with the related art.

According to a third aspect of the present invention, in the code search table, the index having a short code length is set to “0”.
And the value of the minimum Huffman code word of the first group of the code search table given as the initial value is “0”.

By allocating codes such that "0" is given to the shorter code length, it is assumed that the code is zero even if the initial value of the minimum Huffman code word in the group is not given. It can be performed.

[0019]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, embodiments of the decoding device according to the present invention will be described in detail with reference to the drawings.

FIG. 1 is a block diagram showing a configuration of a decoding device 1 according to the present embodiment. The decoding device 1 includes a Huffman table 10, a sample number (Sam_num) register 1
1, code length (length) register 12, signal value (bit_data) organizer 13 to be searched, codeword and maximum codeword (code_max) generator 14, code match determiner 1
5, point generator 16, index (ind)
ex) A register 17 is provided.

Signal value (bit_data) organizer 13 to be searched
Sequentially inputs Huffman coded signals as input signals,
This is an input signal organizing means for organizing into a search bit string according to the code length of the group to be searched.

A codeword and a maximum codeword (code
_max) generator 14 generates the maximum Huffman code word in the group from the minimum Huffman code word in the group to be searched and the number of samples, and calculates the code length of the group and the code length of the next group to be searched. The codeword generation unit generates a next group of minimum Huffman codewords from the difference value and the maximum Huffman codeword.

The code coincidence determiner 15 is a comparing means for comparing the maximum Huffman code word in the group generated by the code word generating means with the input signal organized by the input signal organizing means.

The point generator 16 is a position calculating means for calculating the entry position of the decoding index in the group from the difference value of the comparison result by the comparing means.

Each record of the Huffman table 10 is composed of a decoding index, a code length, and the number of samples.
These records are grouped by code length, and the records in the group are arranged in ascending order of the value of the corresponding Huffman code word. FIG. 3 shows a procedure for creating the Huffman table 10, and FIG.

As shown in FIG. 3, first, a Huffman table is created according to a conventional method (Step 31). For example,
A Huffman table as shown in FIG. 5B is created. At this time, if a code is assigned so as to give “0” to the shorter code length, the code “00” having the smallest value is assigned to the index “9” having the highest appearance probability. In addition, codes are assigned so that Huffman code values are consecutive numbers within a group of the same length. At this time, the indices within the group of the same length need not be in the order of the appearance probabilities.

Next, the created Huffman table is rearranged in ascending order of codewords (Step 32), and further rearranged in ascending order of code length (Step 33). Then, as shown in FIG.
Huffman tables are grouped in descending order of code length for each code length, and the records in the group are arranged in ascending order (serial number) of the value of the corresponding Huffman code word.

Further, the number of samples (sam_num) in the group is set to each record (Step 34). That is, the number of samples is set to 1 for a group having a code length of 2 bits, the number of samples is set to 3 for a group having a code length of 3 bits, and the number of samples is set to 3 for a group having a code length of 4 bits. The number of samples is also 3 for the group, and 6 for the group having the code length of 6 bits.

Since the Huffman table according to the present invention does not require a codeword that is a pair of indexes, it is excluded from the table (dotted line in FIG. 4). That is, this Huffman table is composed of an index,
It is composed of a code length (length) and the number of samples (sam_num). Since the field length of the number of samples according to the present invention can be reduced as compared with a conventionally required codeword, the table size can be reduced as a result. For example,
In FIG. 4 as an example, the codeword requires a field length of 6 bits, but if the field of the number of samples is 4 bits, “0” to “15” can be expressed.

Next, an example of a processing operation for decoding a Huffman coded signal using the Huffman table will be described with reference to the flowchart of FIG. As an example, it is assumed that an encoded signal of a bit string “1011” is input.

Prior to the processing operation, the minimum Huffman codeword (codeword) in the group and the entry position (point) of the Huffman table are given as initial values (Step 0).
1).

Here, "00" is given to the minimum Huffman codeword (codeword) in the group, and "1" (start_point) indicating the head of the table is given to the entry position (point). As described above, if a code is assigned so as to give “0” to the shorter code length when the Huffman table is created, the initial value of the minimum Huffman codeword (codeword) in the group can be omitted. , Zero, the processing can be performed.

Next, the code length (length) and the number of samples (sam_num) indicated by the entry position (point)
Is obtained (Step 02). Here, the entry position (poin
Since t) indicates the head of the table, that is, the index “9”, the corresponding code length “2” and the number of samples “1” are extracted, and the code length register 12 and the code length register 12, respectively.
It is stored in the sample number register 11.

Next, the signal value composer 13 extracts bits of the code length (length) from the beginning of the encoded data as the input signal, organizes them as a search target bit string (bit_data), and Send it (Step03). Here, since the code length (length) of the code length register 12 is “2”, the search target bit string (bit_data) is “10”.

Next, the codeword organizer 14 calculates the maximum Huffman codeword (code_max) in the group by the following equation and sends it to the code coincidence determiner 15 (Step 04).

[0036]

[Equation 1] code_max = codeword + sam_num-1 Here, the minimum Huffman codeword (codeword) in the group
Is “00” and the number of samples (sam_num) is “1”, so that the maximum Huffman codeword (code_max) in the group is “00”.

Next, the calculated maximum Huffman code word in the group (code_max) is compared with the bit string to be searched (bit_data) by the code match determination unit 15 (Step 05). That is,
This is a process of determining whether the target Huffman code exists in the current group.

The maximum Huffman code word (code_ma
x) plus 1 to the search target bit string (bit_dat
If a) is smaller, the target Huffman code is present in the current group, so that the processing of Step 06 to Step 07 is performed to obtain an index value corresponding to the target Huffman code.

Conversely, a bit string (bit_da
When ta) is larger, since it is present in the next and subsequent groups, the processing of Step 11 to Step 14 is performed, and the processing of jumping to the next group is performed.

Here, since code_max + 1 <bit_data, to jump to the next group, the point generator 16 adds the number of samples (sam_num) to the entry position (point) as shown in the following equation ( Step11). Then, “2” is set as the entry position (point).

[0041]

[Formula 2] point = point + sam_num Next, the code length (length) and the number of samples (sam_num) indicated by the entry position (point) are obtained (S
tep12). Here, since the entry position (point) indicates the index “13”, the code length “3” and the number of samples “3” corresponding thereto are extracted and stored in the code length register 12 and the sample number register 11, respectively. You.

Next, the code length (length) of the current group
And the difference between the code length (length) of the previous group (1_dif
f) is obtained (Step 13). Here, the difference value is “1”.

Next, from the obtained difference value (1_diff) and the previous maximum Huffman codeword in the group (code_max), the current minimum Huffman codeword in the group (codeword) is obtained by the following equation (Step 14).

[0044]

[Equation 3] codeword = (code_max + 1) << 1_diff That is, the last maximum Huffman codeword in the group (code_m
ax) Add 1 to “00” and add it to the difference value (1_di
ff) is shifted right by the number of bits. as a result,
The minimum Huffman codeword (codeword) "0
10 "is obtained.

Returning again to Step 03, bits corresponding to the code length (length) are extracted from the head of the encoded data as the input signal and organized as a search target bit string (bit_data). Here, the code length (length) of the code length register 12 is “3”.
, The search target bit string (bit_data) is “101”.

Next, the maximum Huffman code word (code_max) in the group is calculated (Step 04).

Here, the minimum Huffman codeword (codeword) in the group is “010” and the number of samples (sam_num) is
Is “3”, the maximum Huffman codeword (code_max) in the group is “100”.

Next, the calculated maximum Huffman code word in the group (code_max) is compared with the search target bit string (bit_data) (Step 05).

Here, since code_max + 1 <bit_data, the number of samples (sam_num) is added to the entry position (point) in order to jump to the next group (St.
ep11). Then, “5” is set as the entry position (point).

Next, the code length (length) and the number of samples (sam_num) indicated by the entry position (point)
Is obtained (Step 12). Here, the entry position (poin
Since t) indicates the index “1”, the corresponding code length “4” and the number of samples “3” are extracted.

Next, the code length (length) of the current group
And the difference between the code length (length) of the previous group (1_dif
f) is obtained (Step 13). Here, the difference value is “1”.

Next, the current minimum group Huffman codeword (codeword) is calculated from the obtained difference value (1_diff) and the previous maximum group Huffman codeword (code_max) (St).
ep14). The largest Huffman codeword in the previous group (code_m
ax) Add 1 to “100” and further add it to the difference value (1_
diff) bits to the right. As a result, the minimum Huffman codeword (10) in this group is “10”.
10 "is obtained.

Returning again to Step 03, bits of the code length (length) are extracted from the beginning of the encoded data as the input signal and organized as a search target bit string (bit_data). Here, the code length (length) has been changed to “4”,
The search target bit string (bit_data) is “1011”.

Next, the maximum Huffman code word (code_max) in the group is calculated (Step 04).

Here, the minimum Huffman codeword (codeword) in the group is “1010” and the number of samples (sam_nu
Since m) is “3”, the maximum Huffman codeword (code_max) in the group is “1100”.

Next, when the calculated maximum Huffman code word (code_max) in the group is compared with the search target bit string (bit_data) (Step 05), since code_max + 1> bit_data, Jump to the position of the target code (Step 06).

[0057]

[Formula 4] diff = bit_data-codeword point = point + diff Here, "1" is set in diff, and therefore "6" is set in the entry position (point).

Next, an index value “7” indicated by the entry position (point) is obtained and stored in the index register (Step 07), and the process is terminated.

As described above, the decoding apparatus according to the present invention utilizes the fact that code values in a group of the same length are serialized when a Huffman table is created.
A Huffman codeword can be generated from the code length, the number of samples, the minimum codeword in the group that is the initial value, and the like.
Compared with the Huffman codeword required in the conventional Huffman table, the field length of the number of samples required in the present invention can be shortened, and consequently the table capacity can be reduced.

Further, using the generated Huffman code word, the number of samples, and the like, it is determined whether or not the search target bit string exists in the currently searched group, and if so, which entry in the group is located. It is calculated, and if it does not exist, the next group is searched. Therefore, the maximum number of searches can be set to (the number of groups + 1) times, and the processing time required for decoding can be reduced as compared with the related art.

Although the embodiments of the present invention have been described in detail, the present invention is not limited to the embodiments, and various improvements and modifications may be made without departing from the gist of the present invention. .

Accordingly, the present invention should be limited only by the matters specifying the invention according to the claims that are reasonable from this disclosure.

[0063]

As described above, the decoding apparatus according to the present invention utilizes the fact that code values in a group of the same length are serialized when a Huffman table is created.
A Huffman codeword can be generated from the code length, the number of samples, the minimum codeword in the group that is the initial value, and the like.
Compared to the Huffman codeword required in the conventional Huffman table, the number of samples required in the present invention can shorten the field length, resulting in a reduction in table capacity and memory saving. can do.

Also, using the generated Huffman code word, the number of samples, and the like, it is determined whether or not the search target bit string exists in the currently searched group, and if so, which entry in the group is located. It is calculated, and if it does not exist, the next group is searched. Therefore, the maximum number of searches can be set to (the number of groups + 1) times, and the processing time required for decoding can be reduced as compared with the related art.

[Brief description of the drawings]

FIG. 1 is a block diagram showing an embodiment of a decoding device according to the present invention.

FIG. 2 is a flowchart illustrating an example of an operation procedure of a decoding process in the decoding device illustrated in FIG. 1;

FIG. 3 is a flowchart illustrating an example of a procedure for creating a Huffman table used in the decoding device illustrated in FIG. 1;

FIG. 4 is an explanatory diagram showing an example of creating a Huffman table used in the decoding device shown in FIG. 1;

FIG. 5 is an explanatory diagram showing an example of creating a conventional Huffman table.

[Explanation of symbols]

 1. Decoding device 10. Huffman table 11. Sample number register 12. Code length register 13. Signal value organizer to be searched 14. Code word generation 15. Code matcher 16. Point generator 17. Index register

Claims (3)

[Claims] 1. A decoding device for decoding an input Huffman coded signal, wherein each record is composed of a decoding index, a code length, and the number of samples, wherein the records are grouped by the code length, and Records in the group are arranged in ascending order of the value of the corresponding Huffman code word, a code search table in which the minimum Huffman code word of the first group is given as an initial value, and the Huffman coded signal is sequentially input, and the An input signal organizing means for organizing into a search bit string according to a code length; a maximum Huffman code word in the group from the minimum Huffman code word in the group to be searched and the number of samples; Generating the minimum Huffman code word of the next group from the difference value of the code length of the group to be searched and the maximum Huffman code word Codeword generating means, and comparing means for comparing the generated maximum Huffman codeword in the group with the organized input signal; and a position for calculating an entry position of a decoding index in the group from a difference value based on the comparison result. A decoding device, comprising: calculation means. 2. The decoding device according to claim 1, wherein the codeword generating unit adds 1 to the maximum Huffman codeword, and further shifts the codeword to the right by the number of bits according to the difference value. And a next group of minimum Huffman codewords. 3. The decoding device according to claim 1, wherein the code search table is created by an encoding method that gives “0” to an index having a short code length, and the code search table is used as the initial value. A decoding apparatus, wherein a value of a minimum Huffman code word of a leading group of the code search table provided is "0".