Districted matching approach for 1D object classification

This paper proves that the districted matching scheme is more stable than undistricted matching scheme for pattern classification applications, where an object to be classified consists of elements lying on a limited line segment in 1D space. The theoretical result suggests the using of districted matching schemes for pattern recognition/recognition of 1D objects. The method is used in the predication of start codons of nucleotide sequences by artificial neural network based approaches.