Coded data generation or conversion – Digital code to digital code converters – To or from number of pulses
Reexamination Certificate
1999-11-16
2002-02-05
Wamsley, Patrick (Department: 2819)
Coded data generation or conversion
Digital code to digital code converters
To or from number of pulses
C704S501000
Reexamination Certificate
active
06344808
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention relates to an MPEG (Moving Picture Experts Group)-1 audio layer III decoding device, and particularly an MPEG-1 audio layer III capable of fast decoding processing.
2. Description of the Background Art
The MPEG audio is ISO/IEC (International Organization for Standardization and International Electrotechnical Commission) standards of stereo-audio coding with high quality and high efficiency, and is standardized in parallel with coding of motion pictures in MPEG. As a result, products relating to MPEG audio products have recently been developed.
A bit stream of the MPEG audio is formed of frames. Each frame is a minimum unit of data which allows decoding to an audio signal by itself, and always contains a constant sample number of data.
Referring to
FIG. 1
, a frame is formed of a header
1
, an error check
2
, audio data
3
and an ancillary data (external data)
4
. Header
1
is a bit stream portion including information such as a synchronization word, a sampling frequency, a bit rate and others. Error check
2
is optional data, and is a bit stream portion including information for error detection. Audio data
3
is a bit stream portion including information related to an audio sample. Ancillary data
4
is a portion into which data other than MPEG audio can be arbitrarily inserted.
In the MPEG-1 audio layer III, which will be referred to as “MP3” hereinafter, each frame includes 1152 sample data. Each frame including 1152 samples is divided into two granules each including 576 samples.
A breakdown of the 576 samples is as follows. For coding of MP3, an input signal is divided by a sub-band analyzing filter bank into 32 frequency bands based on time regions. The output of each band sent from the filter bank is 18 samples (this kind of output will also be represented as a “a long block” hereinafter), or 6×3 samples (this kind of output will also be referred to as a “short block” hereinafter). Then, each band is mapped to subdivided spectral lines by Modified Discrete Cosine Transform (which will be referred to as “MDCT” hereinafter). Frequency region samples of 18 or 6×3=18 are obtained as the output of MDCT. Accordingly, the frequency resolution is equal to 32 bands×18 samples=576 samples (each sample has an arbitrary data length).
Referring to
FIG. 2
, a decoder of audio data portion
3
in MP3 includes a bit stream decomposing portion
5
which decomposes the input bit string into side information
6
including bit allocation information and Huffman table information, a scale factor and Huffman code data, a scale factor decoder
7
which is connected to bit stream decomposing portion
5
, and decodes the scale factor decomposed from the bit-stream based on side information
6
, a Huffman decoder
8
which is connected to bit stream decomposing portion
5
, and decodes Huffman code data decomposed from the bit-stream based on Huffman table information included in side information
6
, an inverse quantizer
9
which is connected to bit stream decomposing portion
5
, scale factor decoder
7
and Huffman decoder
8
, and performs inverse quantization of the Huffman code data based on side information
6
, scale factor and Huffman code data, and a HFB (Hybrid Filter Bank)
10
which is connected to inverse quantizer
9
and inversely maps the output of inverse quantizer
9
to reconstruct a time region signal.
HFB
10
includes a butterfly operating portion
11
which is connected to inverse quantizer
9
, and conducts a butterfly operation or computation on the inversely quantized signal issued from inverse quantizer
9
, an IMDCT operating portion
12
which is connected to butterfly operating portion
11
, and conducts inverse MDCT (which will be referred to as “IMDCT” hereinafter) on the operation result of butterfly operating portion
11
, and a sub-band composing portion
13
which is connected to IMDCT operating portion
12
, and conducts sub-band composing processing on the operation result of IMDCT operating portion
12
using Polyphase Filter Bank (which will be referred to as “PFB” hereinafter).
Referring to
FIG. 3
, respective portions of the decoder for audio data portion
3
of MP3 operate as follows. In the following description, lengths of data to be handled are not restricted. Bit stream decomposing portion
5
extracts and analyzes header
1
of the received bit string (S
14
). Bit stream decomposing portion
5
decodes the side information
6
, and extracts the Huffman code data and scale factor decomposed from the bit-stream (S
15
). Scale factor decoder
7
decodes the scale factor decomposed from the bit-stream based on side information
6
(S
16
). Huffman decoder
8
decodes the Huffman code data decomposed from the bit-stream based on the Huffman table information included in side information
6
(S
17
). As a result of decoding of the Huffman code data, the Huffman code data of
576
in number are decoded per granule. This number depends on the frequency resolution power.
Inverse quantizer
9
performs the inverse quantization of the Huffman code data based on side information
6
, scale factor and Huffman code data (S
18
). The inverse quantization of the Huffman code data is performed according to the following formula (1):
Xr(i, j)=sign(is(i, j)×is(i, j)×2
P
, 0≦i≦31, 0≦j≦17 (1)
where Xr(i, j) represents a result of the inverse quantization, is(i, j) represents the Huffman code data, P represents a constant obtained from side information
6
and the scale factor, and sign(a) represents the sign of “a”. Further, i represents the subband band number, and j represents the sample number of each subband output.
Butterfly operating portion
11
included in HFB
10
performs the butterfly operation between the sample data of 32 bands issued from inverse quantizer
9
, using 8 samples near the band boundary at a time (S
19
). More specifically, the operation is performed according to the following formula (2):
&AutoLeftMatch;
for
(
i
=
0
;
i
<
31
;
i
++
)
for
(
j
=
0
;
j
<
8
;
j
++
)
{
X
(
i
,
17
-
j
)
=
Xr
(
i
,
17
-
j
)
Cs
(
j
)
-
Xr
(
i
+
1
,
j
)
Ca
(
j
)
;
X
(
i
+
1
,
j
)
=
Xr
(
i
+
1
,
j
)
Cs
(
j
)
+
Xr
(
i
,
17
-
j
)
Ca
(
j
)
;
}
(
2
)
In the formula (2), X(i, j) represents a result of the butterfly operation, and Cs(j) and Ca(j) represent constants determined for the sample numbers, respectively. As to X(i, i) not computed in formula (2), Xr(i, j) is substituted.
IMDCT operating portion
12
obtains data of N samples by folding back the coded N/2 samples, and then performs the inverse transformation (S
20
). The IMDCT processing for the sample band number i is performed according to the following formulas (3) and (4):
Z
i
(
n
)
=
∑
k
=
0
N
/
2
-
1
X
(
i
,
k
)
*
C
(
n
,
k
)
,
0
≤
n
≤
N
-
1
where
(
3
)
C
(
n
,
k
)
=
cos
(
π
2
N
(
2
n
+
1
+
N
2
)
(
2
k
+
1
)
)
(
4
)
where Z
i
(n) represents the intermediate result. N is a constant, and is equal to 36 in the case of the long block. In the case of the short block, N is equal to 12. In the processing at step S
20
, windowing processing is simultaneously performed according to the following formula (5):
H
i
(n)=Z
i
(n)×W(n) (5)
where H
i
(n) represents the result of window operation, and W(n) represents a window coefficient. Processing for overlapped portions is conducted on H
i
(n), whereby the outputs of 18 samples Y(i, j) (0≦j≦17) are obtained as the final outputs with respect to each band i. Hereinafter, the output of IMDCT operating portion
12
with respect to X(i, j) (0≦i≦31, 0≦j≦17) is defined as Y(i, j).
Subband composing portion
13
performs subband composition using PFB, and issues PCM (Pulse Code Modulation) data (i.e., reproduced sample data) which i
Sakamoto Tadashi
Taruki Maiko
McDermott & Will & Emery
Mitsubishi Denki & Kabushiki Kaisha
Wamsley Patrick
LandOfFree
MPEG-1 audio layer III decoding device achieving fast... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with MPEG-1 audio layer III decoding device achieving fast..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and MPEG-1 audio layer III decoding device achieving fast... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2953246