An invariant transform under cyclic permutations for pattern recognition

Abstract EN

A translation invariant transform (A-Transform) for pattern recognition is presented.

This transform maps a sequence of input data into a set of coefficients which are invariant under translation of the input sequence.

Besides the translation invariant property, the proposed transform can be applied on real sequences of any length n.

The transform computation involves addition and substration only, and its complexity is of 0(n2).

Simulations tests show that the A-Transform is superior to the R-Transform and the M-Transform in distinguishing binary and real data of patterns.