Research and application of non-negative matrix factorization with sparseness constraint in recognition of traditional Chinese medicine pulse condition

In this paper, the recognition method based on non-negative matrix factorization with sparseness constraint (NMFs) combined with the support vector machine (SVM) was proposed to identify the type of the common pulse condition of Chinese Traditional Medicine (TCM). First, pulse data were factorized by NMFs to obtain projection coefficients as training sample set to build recognition mode with SVM. Then the method proposed was compared with the classical time-domain method of pulse feature extraction. And time-domain features were extracted to identify the type of pulse with the same SVM classifier. Finally, the results showed that projection coefficients obtained by NMFs more use of recognition of TCM pulse.