Classification of signals by using normal orthogonal transformation

Possibility of pattern recognition is considered on the procedure of creation of discrete ortogonal transformation offered by an author, in that the first transform coincides with a etalon signal. At the coincidence of the investigated signal with a test the spectrum of such transformation contains one unzero transform only, while appearance of other transforms in a spectrum testifies to their differences. Application of normal transformation for the estimation of similarity of signals by coefficients of transforms allows to enter numeral measure of estimation of such similarity. Procedure of recognition is widespread on the cases of two-dimensional and complex signals. Results over of the use of normal transformation are brought for the tasks of authentification of person by the dinamically entered signature and classification of the state of person by the pulse wave.