Method of formulating attributes on the basis of a structural model of the signal

Abstract

In solving acoustic-recognition problems, a special role is played by echo-signal time distortions of the form t → Dt, D > 0. The stability of attribute descriptions relative to such transformations of the time scale is achieved by preliminary measurement of the coefficient D and its subsequent compensation. The latter factor substantially complicates the engineering solution of the system for formulating attributes and significantly increases the duration of the recognition process (see [1]).

The present work proposes a signal-handling method based not on measurement but on the calculation of the time-scale coefficient D according to certain numerical characteristics of the structural representation of the reflected signal. Knowing the scale factor, one can ensure formulation of a stable attribute description of the investigated signal. Moreover, it has been demonstrated that for certain requirements governing the spectrum of the probing signal there exists the possibility of obtaining the same result while avoiding calculation of the coefficient D. This latter situation is realizable due to the implementation of the central idea expounded in [4].