Fast architectures for pattern recognition and classifier training

A systolic architecture is presented for the classification of patterns using the maximum likelihood algorithm. A unique feature of the architecture is the capability of online training. By use of recursive expressions, the classification functions can be updated taking into account all the previous parameters. The proposed architecture is capable of classifying an N-dimensional pattern in (N+1)/2 cycles. It features full use of symmetry properties to speed up computations and to reduce storage requirements