RECOGNITION BY IMITATING THE PROCESS OF PATTERN GENERATION

Publisher Summary This chapter focuses on the pattern recognition problem. There are two main difficulties in the field of recognition. The first difficulty originates from a great variety of the observed signals, while the second difficulty arises from the presence of interference. Most methods of recognition proposed until present are designed to overcome only one of them. The recognition method reviewed in this chapter combines efficient noise-proofness with great versatility. The noise-proofness is insured because of calculation of the resemblance carried out instead of checking the exact coincidence of the observed signal with one of the references. High versatility is achieved by defining a vast number of various reference patterns, which nevertheless are not to be completely scanned to find the most resembling one. The formal statement of the problem enables the judgment of the range of the method applicability. By the approach of character recognition, a problem of separating a typed line without distinct blanks between characters has been solved. T.K. Vintsjuk has achieved high probability in spoken word recognition.