In this paper a new recognition system of distorted patterns is presented, where the Viterbi algorithm and a modified trellis incorporating a pertinent statistics of distorted patterns are used. The system works in a manner quite similar to our own human decision process. The trellis is so constructed that it can eliminate all the irrelevant pattern classes at the outset and leave only the most probable for its fimal decision, yielding a great economy of processing time and the accuracy of decision. The method was applied to recognition of handwritten English letters and Japanese Katakana's with success. The method is also universal and can be applied to any category of distorted pattems.

A new method of soft-decision decoding of threshold decodable codes is presented. Excellent error performance, comparable to or better than that of other known practical schemes, is demonstrated by computer simulation and error analysis. The method is an approximate realization, without undue deterioration in performance, of the optimum {\em a posteriori} probability (APP) decoding and has a much simpler structure.

Decision and estimation algorithms for a signal of asynchronous generation are treated with emphasis on waveform estimation. A simplified model for source statistics is assumed. Starting with a supervised learning model, two algorithms of the non-supervised type are derived. Their estimation error performance is studied by computer simulation.

A noiseless encoding scheme is presented for sources with memory. The source is modeled by an ruth-order finite ergodic Markov process. A secondary source is derived which has as its symbol set the ordinal numbers of the symbol probabilities. The secondary source is approximately memoryless and has entropy smaller than that of the adjoint