Signal classification by mutual information based discriminating pursuit
Bao Liu, Shih‐Fu Ling · 2002
An algorithm called mutual information based discriminating pursuit is suggested for signal classification. With this algorithm a series of wavelets that carry meaningful information about class differences are chosen from wavelet packets. The coefficients obtained by decomposing signals onto these wavelets are exploited as pattern features for classification. The algorithm is tested with the classification of simulation signals. The results show that both the sensitivity and reliability of the proposed algorithm are good. This algorithm can be used to deal with 1-D signal classification, such as speech analysis, machinery diagnosis, etc.