Pseudo-inverse with increasing threshold: an error-recovery patternrecognition algorithm

摘要

A pseudo-inverse algorithm with an increasing threshold value forpattern recognition is presented. This method is particularly usefulwhen the original matrix is not a square matrix. The algorithm and itsimplementation are studied. Using English capital letters as samples,the error-recovery capability of the proposed algorithm is observed tobe better than that of the fixed threshold algorithm. Noisy patternswith 20% errors, i.e., 12 in Hamming distance, can be recognized usingthis approach