JP3501246B2 - MPEG audio decoder - Google Patents

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an MPEG audio decoder for decoding audio data compressed by MPEG.

[0002]

2. Description of the Related Art In recent years, MPEG (Moving Picture Image)
Coding Experts Group) A technique for compressing a digital audio signal called audio has been developed. Current,
MPEG1 is standardized for this MPEG audio. In the MPEG1 standard, three systems of layer I, layer II, and layer III are prepared depending on the required sound quality and circuit scale.

Further, in MPEG audio, each mode such as single channel, dual channel, stereo, and joint stereo is defined as a channel mode of an audio signal. Further, in MPEG audio, various transmission rates (bit rates) that can be used for each channel mode of each layer and output rates (sampling frequencies) of audio signals after reproduction are variously specified.

For example, in the layer II stereo mode,
The audio data is divided into a maximum of 30 subbands, and three pieces of sample information are collectively coded for each subband, and each subband includes two channels.

FIG. 4 is a diagram showing a bit stream of audio data compressed by the layer II of MPEG1. As shown in FIG. 4, one frame of compressed audio data includes header information (HEADER), CRC (Cy
clic Redundancy Check) information, allocation (Allo
cation) information, scfsi information, scale factor (Scale
Factor) information and sample or sample code (Sample / Sample Code) information (hereinafter, these are collectively referred to as sample data), and are input to the MPEG audio decoder in this order.

Here, the header information is composed of basic information such as a bit rate, a sampling frequency, a stereo / monaural channel mode, and the like. The CRC information is for checking whether or not there is an error in the bitstream.

The allocation information, scale factor information, and sample data are the information that constitutes the audio compressed data body. By processing the sample data with the allocation information and the scale factor information, the compressed audio data is decoded and the PCM data is generated.

That is, the allocation information is information used in the dequantization operation at the time of decoding, and information about how many bit widths of the sample data to be input later and the sample data are the sample information. It includes information as to whether it is sample code information. This information is based on ISO / IEC 11172 MPEG.
According to the standard of 1 audio layer II, it can be known by referring to the table information as shown in Table 1 based on the value of the allocation information and the subband number.

[0009]

[Table 1]

In Table 1, the item indicated by nbal indicates the bit width of the allocation information. When the allocation information has a 4-bit width (subbands sb0 to sb0).
In sb10), the allocation information has 16 values of 0 to 15. When the allocation information has a 3-bit width (subbands sb11 to sb22), the allocation information has eight values 0 to 7. Further, when the allocation information has a 2-bit width (subbands sb23 to sb23
In b29), the allocation information takes four values 0 to 3.

For example, when the subband number is sb = 2 and the allocation information value is 5, it can be seen that there are 63 possible values for the sample data and the bit width of the sample data is 6 bits. . Further, when the subband number is sb = 20 and the value of the allocation information is 5, it can be seen that there are 15 possible values for the sample data and the bit width of the sample data is 4 bits.

When the sample data has three possible values, five values, or nine values, the sample data is input to the data input unit 1 in the form of a sample code. Therefore, by referring to Table 1 and seeing how many values the sample data can have,
It is also possible to know whether the sample data is sample information or sample code information.

Here, the sample code is a collection of three pieces of sample information existing continuously in the time axis direction in each subband. For example,
If there are three possible values for one sample information,
Normally, three pieces of sample information have 6 bits, but if combined into one code, 5 bits will suffice. Also, when one sample information has 5 or 9 possible values, normally 3 pieces of sample information each have 9 bits and 1 bit, respectively.
Although it is 2 bits, if combined into one code, 7 bits and 10 bits are sufficient respectively.

The scale factor information is information used in the inverse scaling operation after the inverse quantization operation, and represents the scaling coefficient value that gives an approximate output level. Here, in the case of layer II, there are 36 pieces of the above-mentioned sample information in the time axis direction for each channel of each subband in the case of one frame. Then, the above 36 pieces of sample information are consecutive 1
It is possible to take different scale factor values for every two.

That is, assuming that the scale factor information for the 0th to 11th, 12th to 23rd, and 24th to 35th sample information in the time axis direction is scale factor 0, scale factor 1, and scale factor 2, respectively, As shown in Table 2 of FIG.
Is given a scale factor value of any one of a, b, and c.

[0016]

[Table 2]

The scfsi information is a parameter indicating how many of the above scale factor information exists in each channel of each subband, and is used only in layer II. That is, the scfsi information indicates which one of the four patterns shown in Table 2 corresponds.

The decoding process of the compressed audio data having the bit stream as described above is performed as follows. That is, first, the subband sample information S is obtained by the above-described inverse quantization processing and inverse scaling processing. Next, this subband sample information S
PCM data is generated by performing a synthetic subband filter process using the.

The above-mentioned subband sample information S is obtained according to the following arithmetic expression. S = factor × C × (S ″ + D) (Equation 1) Here, factor in (Equation 1) is a calculated value specified according to the value of the scale factor information, and the scale factor information described above. The value of index
When expressed by, it is given by: Factor = 2 ^{1-index / 3} (Equation 2). Table 3 shows the relationship between the value index of the scale factor information and the value of the above factor.

[0020]

[Table 3]

Further, the value S ″ in the above (formula 1) is an inversion of the most significant bit (MSB) of the sample information. Similarly, the values C and D in the above (formula 1) are predetermined values. The coefficients are the values shown in Table 4. As can be seen from Table 4, different values are used for the coefficients C and D depending on how many values the sample information can have. That is, one of the coefficient values shown in Table 4 is used according to the value of the allocation information.

[0022]

[Table 4]

In the calculation for obtaining the subband sample information S shown in the above (formula 1), first, C × (S ″ + D)
The inverse quantization operation shown by is performed, and then the inverse scaling operation for multiplying the value of the factor is performed. The values of the above factors and the values of the coefficients C and D are those stored in advance as table information as shown in Tables 3 and 4, respectively.

[0024]

However, as is clear from (Equation 1) above, in the conventional MPEG audio decoder, in order to obtain the subband sample information S, an inverse quantization operation and an inverse scaling operation are performed. It is necessary to perform the above two processes, and it is necessary to perform the multiplication process in each process. In general, since the multiplication process has a heavy calculation load, the conventional method of performing the multiplication process twice has a problem that the calculation load becomes large as a whole and the processing time becomes long.

The present invention has been made to solve such a problem, and it is possible to reduce the calculation load when obtaining subband sample information in a series of decoding processes. With the goal.

[0026]

An MPEG audio decoder of the present invention is an MPEG audio decoder adapted to perform inverse quantization processing and inverse scaling processing on MPEG standard audio compressed data,
An operation that separates the exponent of the operation value specified by the value of the scale factor information in the audio compressed data into an integer part and a fractional part, and generates an operation expression for performing the inverse quantization process and the inverse scaling process. Based on the expression generating means, the arithmetic expression generated by the arithmetic expression generating means, the multiplication and addition means for performing arithmetic operations other than the arithmetic operation on the integer part by multiplication and addition, and the arithmetic result performed by the multiplication and addition means. And a shift means for performing a calculation on the integer part in the calculation formula by a shift process.

Another feature of the present invention is that the multiplication and addition means adds the sample information in the audio compressed data and a predetermined coefficient, the addition result by the addition means, and the fractional part. And a predetermined calculation value for the fractional part, which is stored in advance in the storage means as table information.

Another feature of the present invention is that the inverse quantization processing and the inverse scaling processing based on the above-mentioned arithmetic expression are performed by pipeline processing.

Another feature of the present invention is that the first storage means for temporarily storing the sample information, the second storage means for storing the predetermined coefficient in advance, and the predetermined operation value in advance. A third storage means for storing is provided separately, and the adding means is provided at a stage subsequent to the reading means for reading the sample information and the predetermined coefficient, and the sample information and the predetermined information by the reading means are provided. Coefficient reading process,
The addition processing of the sample information and the predetermined coefficient by the addition means is performed in one machine cycle.

[0030]

Since the present invention comprises the above-mentioned technical means, the dequantization process and the descaling process performed in the series of decoding processes are the exponents of the operation values specified by the value of the scale factor information in the audio compression data. Based on an arithmetic expression that is separated into an integer part and a fractional part, it will be performed by being divided into an arithmetic process related to the integer part and an arithmetic process other than the integer part. Since it is possible to process by using a shift operation instead of multiplication, it is possible to reduce the number of times of multiplication that requires a large operation load in the inverse quantization processing and the inverse scaling processing, as compared with the related art.

Further, according to another feature of the present invention, when the inverse quantization process and the inverse scaling process are performed by the pipeline process, the sample information and the predetermined coefficient are read out and the respective information is added. When the processing and the processing are performed in one machine cycle, the shift processing with a small calculation load can be performed in an extremely short time, and the multiplication processing by the multiplication means and the shift processing by the shift means are performed in one machine cycle. Therefore, the above-mentioned read processing and addition processing, and the multiplication processing and shift processing using the information obtained by the read processing and addition processing in the previous machine cycle,
All can be done in one machine cycle.

[0032]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of an MPEG audio decoder of the present invention will be described below with reference to the drawings. FIG. 1 is a claim correspondence diagram showing a configuration of an MPEG audio decoder according to the present invention.

In FIG. 1, is an arithmetic expression generating means, which is an arithmetic expression for performing inverse quantization processing and inverse scaling processing on audio compressed data of the MPEG standard. It is transformed to generate a new arithmetic expression. That is, the exponent of the operation value specified by the value of the scale factor information included in the audio compression data from an external medium (not shown) such as a compact disc is divided into an integer part and a fractional part, and a conventional expression is obtained. It is transformed to generate an arithmetic expression for performing the inverse quantization process and the inverse scaling process.

Further, is a multiplying / adding means, and based on the arithmetic expression generated by the arithmetic expression generating means, arithmetic operations other than the arithmetic operation on the integer part are performed on the input audio compressed data by the multiplication / addition. Is a shift means, and the sub-band sample information S is obtained by performing a shift process on the result of the arithmetic operation performed by the multiplying and adding means with respect to the integer part in the arithmetic expression generated by the arithmetic expression generating means. Seeking
Output to an internal memory (not shown).

FIG. 2 is a functional block diagram showing the structure of an MPEG audio decoder which embodies the present invention shown in FIG. In FIG. 2, reference numeral 1 denotes a data input unit, which is an MPE composed of the bit stream shown in FIG.
Input G standard audio compressed data. As described above, the compressed audio data is input to the data input unit 1 in the order of the bit stream shown in FIG. Inside the data input unit 1, a header register 1a is provided, and the header information first included in the bit stream is stored.

A CRC check unit 2 detects an error in the bit stream based on the CRC information (CRC check word). That is, the CRC check unit 2
Is a 16-bit CRC checkword value, the header information that was input before that, and the allocation information, scfsi information, and scale factor information values that are input after the CRC checkword. The 16-bit value obtained as a result of the comparison is compared. Then, the error detection is performed by checking whether or not the values of the both match.

3 is an AS-RAM (Allocation / Scale Fa
ctor-RAM) block, which includes a first serial / parallel conversion unit (S / P conversion unit) 3a, a first writing unit 3b, a first reading unit 3c, and an AS-RAM 3d. There is. The first S / P converter 3a
Is for converting the allocation information, scfsi information and scale factor information sent from the data input unit 1 in the form of serial data into the form of parallel data.

The AS-RAM 3d temporarily stores the parallel-converted allocation information, scfsi information, and scale factor information.
Reading and writing are performed by the writing unit 3b and the first reading unit 3c. This AS-RA
In M3d, the scfsi information stored once is overwritten by the scale factor information input later. Therefore, finally, only the allocation information and the scale factor information are stored in the AS-RAM 3d.
Will be stored in.

Reference numeral 4 is a sample RAM block, which is the second
The S / P converter 4a, the sample converter 4b, the second write unit 4c, the sample RAM 4d, the second read unit 4e, and the adder 4f. The second S / P converter 4a converts the sample (or sample code) information sent from the data input unit 1 in the form of serial data into the form of parallel data.

The sample conversion section 4b decomposes the input sample code information and converts it into three pieces of sample information. The sample RAM 4d temporarily stores the parallel-converted sample information, and is read and written by the second writing unit 4c and the second reading unit 4e. The adder 4f adds the sample information read by the second reading unit 4e and the predetermined coefficient D read from the first coefficient ROM 5 by the third reading unit 7.

Reference numeral 8 denotes an arithmetic unit, which includes a first register 9 and
It has a multiplier 10, a shifter 11, and a second register 12. The first register 9 has the adder 4f.
The result of addition in is temporarily stored. The multiplier 10 causes the third reading unit 7 to add the second coefficient R to the addition result temporarily held in the first register 9.
It is multiplied by a predetermined coefficient C ′ read from the OM6. The shifter 11 shifts the multiplication result of the multiplier 10 by a predetermined amount. The second register 12 temporarily holds the subband sample information S obtained as a result of the shift operation in the shifter 11.

The shifter 11 corresponds to the shift means shown in FIG. 1, and the adder 4f and the multiplier 10 correspond to the multiply-add means shown in FIG.

Next, the operation of the MPEG audio decoder configured as described above will be described. First, header information is input from the data input unit 1 and stored in the header register 1 a in the data input unit 1. Next, the CRC information is input from the data input unit 1 and transferred to the CRC check unit 2. Then, the CRC check unit 2 checks whether or not there is a bit error in the frame.

Next, allocation information, scfsi information, and scale factor information are arranged in this order in the data input section 1.
Input from the first S in the form of serial data
The data is sequentially transferred to the / P converter 3a. In the first S / P converter 3a, the allocation information, scfsi information and scale factor information are converted from serial data form to parallel data form. Then, each piece of information thus parallel-converted is stored in the first writing unit 3b.
Stored in the AS-RAM 3d.

As described above, in the stereo mode of Layer II, the audio data is divided into a maximum of 30 subbands, each subband is coded, and each subband includes two channels. Has been. AS- above
The RAM 3d is provided with addresses 0 to 59 for storing the above information for each channel of each subband.

Next, the sample data is input from the data input section 1, and the sample data is sent in the form of serial data to the second S / P
It is transferred to the conversion unit 4a. In the second S / P converter 4a, the sample data is converted from serial data form to parallel data form every 2 bytes.

From the data input section 1 to the second S /
When transferring to the P conversion unit 4a, the number of bits of the transferred sample data is counted, and at the same time AS-
The allocation information of each subband and channel is read out from the RAM 3d in order from a younger address. As a result, it is constantly monitored which subband and channel the sample data corresponding to is being transferred.

When the sample data corresponding to a certain sub-band or channel thus transferred is sample code information (as described in the conventional example, whether or not it is sample code information is determined according to the value of allocation information). However, the parallel-converted sample code information is decomposed into three sample information by the sample conversion unit 4b. 3 from sample code information
The following process is performed to extract one sample information.

[0049] for (i = 0; i <3; i ++) { s [i] = c% nelvels c = c div nelvels } However, s [i] (i = 0 to 2): 3 sample information c: Initial value is sample code information nelvels: 3or5or9 %: Remainder div: quotient

The sample information obtained by the predetermined processing in the second S / P conversion section 4a and the sample conversion section 4b is then given to the second writing section 4c and stored in the sample RAM 4d. .

FIG. 3 is a diagram showing an example of stored contents of the sample information stored in the sample RAM 4d. In FIG. 3, the sample information is shown as sp (i, j, k). Here, i indicates the channel number, j indicates the subband number, and k indicates the order in the time axis direction.

As shown in FIG. 3, the sample information is s
p (0,0,0), sp (0,0,1), sp (0,
0,2), sp (1,0,0), sp (1,0,1),
The data are sequentially stored in the sample RAM 4d in the order of sp (1,0,2), sp (0,1,0), ....

This sample information has a different bit width for each channel of each subband (2 to 16 bits).
However, in the present embodiment, each sample information having a different bit width is not stored in each word one by one, but is stored after the sample information previously stored in succession. . Therefore, the sample RAM 4d stores some sample information in one address. Further, as shown by the hatched portion in FIG. 3, one sample information may be stored across two addresses.

As described above, the sample information is sequentially stored in the sample RAM 4d, and when a certain amount is stored, the decoding process is performed next. Even during this decoding process, the bit stream of the compressed audio data is always input from the data input unit 1 and each information is stored in the AS-RAM 3d and the sample RAM 4d.

In this decoding process, the subband sample information S is first obtained. Conventionally, the above-described subband sample information S has been obtained according to the above-described equations (1) and (2). As shown in Table 3, since the value index of the scale factor information takes a value of 0 to 62, the above (Formula 2) divides the exponent into an integer part m and a fractional part n, and ) Can be transformed into. Factor = 2 ^{1-index / 3} = 2 ^-m (2 ^{-n / 3} ) (Equation 3)

Therefore, the above (formula 1) can be transformed into the following (formula 4) using this (formula 3). S = 2 ^−m · (2 ^{−n / 3} × C) × (S ″ + D) (Equation 4) However, m = −1 to 19, n = 0 to 2 Further, in (Equation 4), 2 ^{−n Supposing that / 3} × C = C ′, S = 2 ^−m · C ′ × (S ″ + D) (Equation 5).

In this embodiment, the sub-band sample information S is obtained by performing the inverse quantization operation and the inverse scaling operation based on this (Equation 5). Here, since the calculation of 2 ^−m in (Equation 5) can be performed by a shift calculation, the subband sample information S is obtained by a calculation process of one addition, one multiplication, and one shift calculation. You can

As is well known, the calculation load of the shift calculation is much smaller than that of the multiplication. Therefore, according to the present embodiment, the multiplication having a large calculation load has to be performed twice. , Subband sample information S
It is possible to significantly reduce the calculation load for obtaining.

This will be described with reference to FIG. First, the sample RA by the second reading unit 4e.
The sample information S ″ is the most significant bit (MSB) from M4d
Are read in the inverted state. At the same time, the first
The sub-band corresponding to the read sample information, channel allocation information, and scale factor information are read from the AS-RAM 3d by the reading unit 3c.

Then, the coefficient C'and the coefficient D corresponding to the values of the read allocation information and scale factor information are read from the first and second coefficient ROMs 5 and 6 by the third reading section 7, respectively. The sample information S ″ and the coefficient D read in this way
Is given to the adder 4f to be added. Then, the addition result (S ″ + D) in the adder 4f is
Is transferred to and held in the first register 9 via.

Next, the multiplier 10 multiplies the addition result (S ″ + D) held in the first register 9 by the coefficient C ′ read by the third reading section 7. Subsequently, the multiplication result {C ′ × (S ″ + D)} in the multiplier 10 is given to the shifter 11 and 2 ^−m
Is shifted by the amount of.

As described above, the sub-band sample information S = 2 ^−m · C ′ × (S ″ + D) is obtained and held in the second register 12. This second register 1
The subband sample information S held in 2 is used for the subsequent synthesis subband filter processing.

Table 5 shown below shows the total number of machine cycles required when the above-described inverse quantization operation and inverse scaling operation are performed by pipeline processing. It should be noted that (a) and (b) in Table 5 show the number of machine cycles when the calculation is performed based on the conventional (Equation 1), and (c) is the same as (Equation 5) of the present embodiment. The number of machine cycles when the calculation is performed based on this is shown.

[0064]

[Table 5]

Table 5 (a) shows a case where the coefficient factor, the coefficient C, and the coefficient D are stored in one coefficient ROM. In this case, first, in the first machine cycle, the value S ″ obtained by inverting the MSB of the sample information and the coefficient D are read out. Then, in the second machine cycle, the coefficient C is read.
And the sample information S ″ read in the first machine cycle and the coefficient D are added.

In the next third machine cycle, the coefficient factor is read and the coefficient C read in the second machine cycle is multiplied by the addition result (S "+ D). , The third above
, The sub-band sample information S is obtained by multiplying the coefficient factor read in the machine cycle of {circle around (1)} and the multiplication result {C × (S ″ + D)}. Since the coefficient D is read out, the subband sample information S is obtained in three machine cycles as a whole.

In Table 5 (b), two coefficient ROMs are provided,
The coefficient D is stored in the first coefficient ROM and the second coefficient R is stored.
The case where the coefficient factor and the coefficient C are stored in the OM is shown. When the coefficient ROM is divided into two in this way, the next sample information S ″ and the coefficient D can be read at the stage of the third machine cycle, so that the subband sample information can be read in two machine cycles as a whole. It becomes possible to obtain S.

In contrast to the conventional example as described above, in (c) of Table 5 showing the operation of the present embodiment, first, the sample information S ″ is read from the sample RAM 4d in the first machine cycle, and the first and second samples are read. The coefficient C ′ and the coefficient D are read from the coefficient ROMs 5 and 6 of No. 2. Further, in the first machine cycle, the processing of adding the read sample information S ″ and the coefficient D in the adder 4f is also performed. .

The reason why the two processes of reading the sample information S ″ and the coefficient D and the adding process thereof can be performed in one machine cycle is as follows: FIG. As shown in, sample R
In AM4d, one sample information is often stored over two addresses.

Therefore, in order to read one piece of sample information, the reading process must be performed at least twice. Therefore, the sample RAM 4d
Until one piece of sample information is output to the bus 13, a combining process using a latch (not shown) is performed. Therefore, in the pipeline processing, it is possible to obtain a sufficient time for adding S ″ + D while the latch processing is being performed.

Next, in the second machine cycle, the coefficient C'read in the first machine cycle and the adder 4f
Multiply the addition result (S ″ + D) in
The multiplication result {C ′ × (S ″ + D)} is subjected to a shift operation by 2 ^−m.Since the shift operation requires a very short operation time as compared with the multiplication, it can be performed in the same machine cycle. It is possible to perform the above multiplication and shift operations.

In the second machine cycle, since the next sample information S ″ and the coefficients C ′ and D are read,
As a whole, the subband sample information S can be obtained in one machine cycle. In this way, the product-sum operation is 1
In the case of using the arithmetic unit 8 capable of processing in a machine cycle, according to the present embodiment, the processing which took at least two machine cycles in the past can be performed in one machine cycle, and the calculation time can be greatly shortened. it can.

In the above embodiment, two blocks, the AS-RAM block 3 and the sample RAM block 4, are provided so that the allocation information and the scale factor information and the sample information are stored in different RAMs. There is. However, the present invention is not limited to this, and the above-mentioned information may be stored in the same RAM.

Although the above description is based on layer II of the MPEG1 audio standard, the present invention is based on layer I or layer III of MPEG1, or MPEG.
The same can be applied to the second standard.

[0075]

As described above, according to the present invention, the exponent of the operation value specified by the value of the scale factor information in the audio compression data is divided into the integer part and the fractional part, and based on the operation formula, Since the inverse quantization process and the inverse scaling process are performed by the multiplication and addition operation and one shift operation, one of the multiplications that has been conventionally performed twice is replaced with the shift operation, thereby significantly reducing the operation load. Therefore, the processing time for performing the inverse quantization process and the inverse scaling process can be shortened.

According to another feature of the present invention, the inverse quantization processing and the inverse scaling processing are performed by pipeline processing, and the reading processing of the sample information and the predetermined coefficient and the addition processing of the respective information are performed. Since the above is performed in one machine cycle, the above-described read processing and addition processing and the multiplication processing and shift processing using the information obtained by the read processing and addition processing in the previous machine cycle are all performed in one machine. This can be performed in cycles, and the processing speed for performing the inverse quantization process and the inverse scaling process can be further increased.

[Brief description of drawings]

FIG. 1 is a claim correspondence diagram showing a configuration of an MPEG audio decoder of the present invention.

FIG. 2 is a functional block diagram showing a configuration of an MPEG audio decoder according to an embodiment of the present invention.

FIG. 3 is a diagram showing an example of stored contents of sample information stored in a sample RAM.

[Fig. 4] Fig. 4 is a diagram illustrating a bit stream of audio compression data according to layer II of MPEG1 audio.

[Explanation of symbols]   Formula generation means   Means of multiplication and addition   Shift means 4 sample RAM block 4d sample RAM 4e Second reading unit 4f adder 5 First coefficient ROM 6 Second coefficient ROM 7 Third reading section 8 arithmetic unit 9 First register 10 multiplier 11 shifters 12 Second register

Front page continuation (72) Inventor Toshiyuki Naoe 2-6-3 Otemachi, Chiyoda-ku, Tokyo Shin Nippon Steel Co., Ltd. (72) Inventor Satoshi Nishimura 2-6-3 Otemachi, Chiyoda-ku, Tokyo Shinichi This steel company (72) Inventor Tsutomu Nonaka 2-6-3 Otemachi, Chiyoda-ku, Tokyo Nippon Steel Co., Ltd. (72) Inventor Naohisa Suzuki 2-6-3 Otemachi, Chiyoda-ku, Tokyo Nippon Steel Co., Ltd. (72) Inventor, Yasho Sato 2-6-3 Otemachi, Chiyoda-ku, Tokyo Nippon Steel Co., Ltd. (56) References Patent 3190204 (JP, B2) (58) Fields surveyed (Int.Cl. ⁷ , DB name) G10L 19/02

Claims (4)

(57) [Claims] 1. An MPEG audio decoder adapted to perform inverse quantization processing and inverse scaling processing on MPEG standard audio compressed data, wherein an arithmetic value specified by a value of scale factor information in the audio compressed data. The exponent of is separated into an integer part and a fractional part, and an arithmetic expression generating means for generating an arithmetic expression for performing the inverse quantization processing and the inverse scaling processing, and an arithmetic expression generated by the arithmetic expression generating means. Based on the above, the multiplication and addition means for performing operations other than the operations related to the integer part by multiplication and addition, and the shift means for performing the operations related to the integer part in the above arithmetic expression by a shift process with respect to the operation result performed by the multiplication and addition means. MP comprising:
EG audio decoder. 2. The multiplying / adding means multiplies an adding means for adding the sample information in the audio compressed data and a predetermined coefficient, and a multiplication for multiplying the addition result of the adding means by a predetermined operation value for the fractional part. 2. The MPEG audio decoder according to claim 1, wherein the MPEG arithmetic decoder is configured by means and means, and the predetermined operation value for the fractional part is stored in the storage means in advance as table information. 3. The M according to claim 1, wherein the inverse quantization processing and the inverse scaling processing based on the arithmetic expression are performed by pipeline processing.
PEG audio decoder. 4. A first storage means for temporarily storing the sample information, and a second storage means for storing the predetermined coefficient in advance.
Storage means and third storage means for storing the predetermined calculation value in advance, respectively, and the adding means is provided in the next stage of the reading means for reading the sample information and the predetermined coefficient. The reading process of the sample information and the predetermined coefficient by the reading unit and the addition process of the sample information and the predetermined coefficient by the adding unit are performed in one machine cycle. Three
MPEG audio decoder described in.