Information compression method

Abstract

Abstract PURPOSE:To compress an information amount by applying an information signal to be processed to a white filter, inputting its output signal to a linear dynamic filter and processing its output at a lattice filter while regarding its output as the information signal to be processed. CONSTITUTION:A signal x(t) of unsteady state probability is formed into an orthogonal signal u(t) by a white filter 1 and the signal is inputted to the linear dynamic system 7. When the autocorrelation function of the signal x(t) with respect to time is known and the signal u(t) is inputted while making system parameters A(t)-C(t) adaptive, a signal y(t) satisfying an output of the system 7 follows the Markov process and its autocorrelation function is nearly equivalent to that of the x(t). A signal z(t) adding a noise w(t) to a signal obtained by multiplying a weight coefficient C(t) to the signal y(t) is given to a lattice filter 6, its output is subject to sampling delay 4, the outputs are added while multiplying coefficients a1-an to each so as to form a signal y'(t). The compression is attained by selecting the coefficients a1-an so that I=y(t)-y'(t) is minimized.

Inventors

• 👤 Toshihiro Furukawa
• 👤 利博 古川

Bibliographic Details