Electrical computers: arithmetic processing and calculating – Electrical digital calculating computer – Particular function performed
Reexamination Certificate
1999-01-28
2001-10-23
Ngo, Chuong Dinh (Department: 2121)
Electrical computers: arithmetic processing and calculating
Electrical digital calculating computer
Particular function performed
Reexamination Certificate
active
06308194
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention relates to a discrete cosine transform circuit which can be used in compression/extension processing of digital voice data in a digital voice recording device and a reproduction device and its operation method, and especially to the enhancement of the processing efficiency and the reduction of the power consumption.
2. Description of The Related Art
FIG. 1
is a block diagram showing a processing device which performs encoding/decoding of digitized voice data. At the time of recording, the entered voice signal is first converted to digital voice data by A/D (analog to digital) converter
2
. The digital voice data is divided to three low, medium and high frequency bandwidths using QMF (quadrature mirror filter) circuit
4
. The digital time series voice data is converted to frequency component data using DCT (discrete cosine transform) circuit
6
, and further quantized by a quantizing unit
8
. The generated or encoded data is supplied to the next-stage processing circuit, and recorded in a predetermined recording medium.
On the other hand, at the time of reproduction, processing reverse to the processing described above is performed. Specifically, an inverse quantizing unit
10
, IDCT (inverse discrete cosine transform) circuit
12
, IQMF (inverse quadrature mirror filter) circuit
14
and D/A (digital to analog) converter
16
perform the conversion reverse to the conversion performed by the quantizing unit
8
, DCT circuit
6
, QMF circuit
4
and A/D converter
2
. Specifically, a voice signal is reproduced from the recorded encoded data.
Additionally, DCT is useful for encoding/decoding voice signals, and it has been widely used. There are various types of DCT. For example, there is one type of DCT for use in a voice recording/reproducing device which is represented by the following relational equation of 2M items of time series voice data y(n) represented by a time index n which is a continuous integer and M items of frequency component data X(k) represented by a wave number index k which is a continuous integer:
y
(
n
)
=
∑
K
=
0
M
-
1
X
(
k
)
cos
(
π
(
2
k
+
1
)
(
2
n
+
M
+
1
)
4
M
)
(
0
≤
n
<
2
M
)
(
1
)
The DCT is slightly modified from a basic DCT and is, therefore, termed a Modified DCT, and will hereinafter be abbreviated as MDCT. Moreover, the inverse modified DCT is hereinafter abbreviated as IMDCT.
As an algorithm for processing DCT at a high rate, a method is known in which FFT (fast Fourier transform) is used. By the algorithm using FFT, sequence y(n) is obtained from sequence X(k) in the MDCT. Conversely, sequence X(k) is obtained from sequence y(n) in IMDCT.
More specifically, the relational equation (1) of the time series voice data y(n) and the frequency component data X(k) is represented in a format suitable for the calculation of IMDCT. For MDCT, calculation0 is performed based on equation (6) described later.
The calculation algorithm regarding IMDCT based on the equation (1) will be described hereinafter. First, the data before conversion, i.e., sequence X(k) is re-arranged and re-constructed according to the predetermined rule to define a new sequence U(k). Based on U(k), Z(j) represented in the following equation is defined. Additionally, in the equation, i denotes an imaginary number unit, and &psgr;(j) denotes the predetermined function of j.
Z(j)=(U(2j)+iU(2j+1)•exp(i&psgr;(j)) (2)
Furthermore, z(n) defined by the following equation is obtained from Z(j).
z
(
n
)
=
∑
j
Z
(
j
)
exp
(
i
Ψ
′
(
j
)
)
(
3
)
In order to calculate the equation (3) at high speed, FFT is used. As is well known, FFT calculates the above equation (3) by repeating the arithmetic operation represented by the following equation. Additionally, &psgr;′(j) is the predetermined function of j.
Z(j
1
)+z(j
2
)•exp(i&psgr;′(j)) (4)
In IMDCT, u(n) defined in the following equation (5) is obtained from the z(n), and the sequence u(n) is re-arranged and re-constructed to obtain the time series voice data y(n). Additionally, ao to a
3
are proportional coefficients defined for every n.
u(n)=a
0
•Rez(n)+a
1
•Rez(M/2−1−n) +a
2
•Imz(n)+a
3
•Imz(M/2−1−n)
u(M−1−n)=a
2
•Rez(n)−a
3
•Rez(M/2−1−n) −a
0
•Imz(n)+a
1
•Imz(M/2−1−n) (5)
On the other hand, for MDCT, the following relational equation is used to obtain the frequency component data X(K) from the sequence x(n) based on the time series voice data y(n).
X
(
k
)
=
2
M
∑
k
=
0
M
-
1
x
1
(
n
)
cos
(
π
(
2
k
+
1
)
(
2
n
+
M
+
1
)
4
M
)
(
6
)
The equations (1) and (6) have substantially the same format except the coefficient 2/M. Therefore, the calculation algorithm of MDCT is expected to be similar to that of the IMDCT described above. In practice, the calculation algorithm of MDCT based on the equation (6) is as follows, and has points common with the IMDCT algorithm.
First, a new sequence x′(n) is defined by the sum (or difference) of the predetermined elements of the data before conversion, i.e., the sequence x(n) as shown in the following equation:
x′(n)=x(n
1
)+x(n
2
) or x(n
1
)−x(n
2
) (7)
Based on the x′(n), z(j) is defined in the following equation having the same format as that of the equation (2):
z(j)=(x′(2j)+ix′(2j+1))•exp(i&psgr;(j)) (8)
Furthermore, Z(k) is obtained from the z(j) as defined in the following equation:
Z
(
k
)
=
∑
j
Z
(
j
)
exp
(
ⅈΨ
′
(
j
)
)
(
9
)
The equation (9) has the same format as that of the equation (3), FFT is also used in the high speed calculation in the same manner as in the equation (3), and the arithmetic operation is performed in the following format:
z(j
1
)+z(j
2
)•exp(i&psgr;′(j)) (10)
In MDCT, the frequency component data X(k) is obtained from the Z(k) by the following equation (11):
X(k)=b
0
•ReZ(k)+b
1
•ReZ(M/2−1−k)
+b
2
•ImZ(k)+b
3
•ImZ(M/2−1−k)
X(M−1−k)=b
2
•ReZ(k)−b
3
•ReZ(M/2−1−k)
−b
0
•ImZ(k)+b
1
•ImZ(M/2−1−k) (11)
In the equation, b
0
to b
3
are proportional coefficients determined for each k. When the proportional coefficient a
L
(L=0 to 3) determined for each n is represented as a
L
=a
L
(n) or the proportional coefficient b
L
(L=0 to 3) is represented as b
L
=b
L
(k), the following relationship is established between the coefficients:
b
L
(j)=a
L
(j)×2/M (12)
FIG. 2
is a block diagram showing a conventional IMDCT circuit in which the aforementioned IMDCT arithmetic operation is realized. The data before conversion, i.e., the frequency component data X(k), is stored in RAM (random access memory)
20
. The RAM
20
is also constituted to store the results during the course of the arithmetic operation. For example, the proportional coefficient a
L
(L=0 to 3) is stored in ROM (read only memory)
22
. The value read from RAM
20
and held in a register
26
and the value read from ROM
22
and held in a register
28
are transmitted to a multiplier
24
, which multiplies these values to transmit them to either register
30
or
32
.
An adder/subtracter
34
has two inputs A, B, which are connected to selectors
36
,
38
. The registers
26
and
30
are connected to the input side of the selector
36
. Therefore, the selector
36
can selectively supply the data stored in RAM
20
or the data multiplied by t
Fuma Masato
Matsui Masaru
Cantor & Colburn LLP
Ngo Chuong Dinh
Sanyo Electric Co,. Ltd.
LandOfFree
Discrete cosine transform circuit and operation method thereof does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Discrete cosine transform circuit and operation method thereof, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Discrete cosine transform circuit and operation method thereof will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2614240