CS212976B1 - Single-purpose analog computer for calculation (51) Int. Cl. 'G 06 G 7/22 Discrete Cosine Transforms - Google Patents

Abstract

The invention relates to the field of communication technology and cybernetics. The invention solves the problem of reducing redundancy and irrelevance of data transmitted by a television telecommunications system. It can also be used to pre-process image information at the input of a recognition device. The essence of the invention is analog computing power, which solves the discrete cosine, or Fourier or Walsh-Hadamard transformation of input data.

Description

The present invention relates to a dedicated analog computer in which discrete cosine transforms of input data are solved using an analog computing network.

Previously known single-purpose computers, which perform cosine transformation of input data, work on the principle of digital, digital information processing. Since these dedicated computers are usually designed to process video information in the online mode, high demands are placed on their speed. Although some of these computers use an algorithm similar to the so-called Pourier transform algorithm, their speed is insufficient.

The above drawback is overcome by a dedicated analog computer for calculating the discrete cosine transform according to the invention, which consists of a series connection of an analog memory block, an analog calculation block and an analog digital converter block, the analog calculation block comprising a large computing network and operational amplifier block. involved in the era.

The advantage of using an analog computing network is that the calculation of discrete cosine transformation is accelerated, so that it is possible to implement a method of transforming data into real-time systems.

Fig. 1 shows a block diagram of a dedicated analog computer according to the invention. Fig. 2 shows the connection of a computer network and operational amplifiers. Fig. 3 shows a block diagram of a 4x2 analog cosine transformation analog calculation block.

The electrical signals f (x ^) to be transformed are applied either directly or via the analog buffer 6 (see FIG. 1) to the input of the analog calculation block X. The block χ comprises a large counting network 200 composed of resistors 11 to 41, See Fig. 2, connected to the input signal buses 50 to 53 and the operational amplifier block 100. The computer network consisting of resistors 10 to 43 is designed so that block X performs discrete cosine or Walsh-Hadamard or S-transformers of input give.

From the operational amplifier block 100, the output signals are routed to the input of the analog-to-analog converter block 2, which converts them into a binary form.

We will now describe in detail the specific case where we require Block 1 to perform a discrete fourth order cosine transformation. For the calculation of one-dimensional spectral coefficients c (0), c (k) of a discrete cosine transformation, the following formulas apply:

c (0) =

n = 0 (f)

N-1 for k (1, 2, N-1)

F (n) cos

2n + 1 k

212 976

If we interpret a one-dimensional discrete cosine cosine transformation of the sequence of N elements of the signal f (n), nt £ o, l, ... N-lj algebraically as a transformation of vector coordinates in N-dimensional metric space, then between the column matrix of vectors , the column matrix of the transformed signal vector ccJ and the transformer core [Cj] applies the relation

Ol W · M ·

This transformation relationship assigns the base vectors of the transformed spectral system for the respective base system vectors. Coordinates of transformation vectors represent directly the elements of the corresponding row of the transformation matrix C ^ (j) of this transformation matrix are calculated according to the formulas:

cos i + 1, 2N where ié [o, l, ... N - 1] is J1, 2, ... N - lj

If we perform a concrete calculation for N vectors of transformation:

C ° (0) cJ (O) cj (o) cJ (O) cj (1) cjd) ojd) cjd) cj (2) cj, (2)

C ² (2) cj (2) cj (3) ^C 4 13 ⁾

4, we get the following coordinate values

0.5

0.5

0.5

0.5

0,653

0.270

-0.270

-0.653

0.5

-0.5

-0.5

0.5

0.270

-0.653

212 976

C ² (3) = 0.653 c (3) = -0.270

The values of resistors 3.0 to 43 of the counting network, Fig. 2, are now chosen so that the relation holds

R ₁₀ = K 'C ° (0) ^R 11 ^K * ^C 4 ⁽⁰⁾ etc., generally ^B ii, 1' ^K · where R 1, is the ohmic value of the resistor having its position in Figure 2, and a suitable proportionality constant ¹ Jj.

If we apply to the input signal buses 50 to 53 of this designed analog voltage calculation block proportional to the signals f (1) to f (4), 60 to 63 voltages appear at the outputs of the operational amplifiers, proportional to the spectral coefficients c (0) to c ( 3).

In a similar way, it is possible to design a computing network for the calculation of the higher order cosine transformation.

It is also possible to design a two-dimensional discrete cosine transform computation network on the same principle. First, the submatrices ^N Cu, v (x, y) are calculated transform transform matrix ^{N N} Cu, v (x, y) 3 for x, y, u, v ^ O 0 í 1-N 1 and the values of submatrix elements are assigned corresponding values of the resistance of the computer network. Upon returning, the cell is assumed that the output signals of the analog computation block, proportional to the spectral coefficients c (u, v), u ≥ 0 - M - 1j, νζ £ θ- ~ -Ν - 1j, will be taken in parallel, ie. the operational amplifiers must be equal to χ χ N, or M amplifiers and a number of N - pole switches of analog signals - see Fig. 3 are shown. This figure shows an analogue calculation block of two - dimensional, cosine transforms of the order of 4x2.

The block comprises the input signal buses 50 to 22 of the small counting network 300 and 400, the analog signal switch group 500, controlled by the control block 501, and the operational amplifier block 100. The control block 501 and the group of analog signal switches provide cyclic connection of the computer network outputs to the operational amplifier inputs. At the time of the first small counting network 300 being connected, signals proportional to the spectral coefficients o (u, 1) are taken from the outputs of the operational amplifiers. 3).

Similarly, computation networks of other transformations, such as Walsh - Hadamard, Fourier, or S - transforms, can also be designed using appropriate formulas.

The invention can be used to digitally encode and reduce the redundancy and irrelevation of video signals in the field of television transmission technology and automation.

Claims (2)

A single-purpose analog computer for discrete cosine transformation calculation, characterized in that it consists of a series connection of an analog memory block (3) of an analog calculation block (1) and an analog-to-digital converter block (2), a network (200) and a block (100) of operational amplifiers connected in series. A single-purpose analog computer for calculating the discrete cosine transform according to claim 1, characterized in that the large computing network in the analogue calculation block (1) consists of at least two small computing networks (300, 400) and small computing networks (300, 400). a group (500) of analogue signal switches connected to their control block (501) by their control inputs.