Lifted wavelet transforms with application to human pulse identification

By means of a commonly used lifting decomposition algorithm based on D4 wavelet, the first-level lifted wavelet transform is applied to the pulse waves obtained from 20 healthy normal subjects and 20 heroin addicts. Then, 2-dimensional feature vectors are extracted that can be used to effectively identify the healthy normal subject and heroin addict pulse waves. On this basis, the authors of this paper carefully researched the algorithm principle of lifting wavelet transforms. Considering the non-uniqueness of polyphase matrix decompositions, the authors derived out another decomposition form of the polyphase matrix based on D4 wavelet, as well as its corresponding lifting realization algorithm. The derived algorithm is, respectively, used to the first-level lifted wavelet transform of the 40 pulse waves. Then, the 2-dimensional feature vector is extracted for every pulse wave. The result of experimental analysis shows that the algorithm presented by this paper, comparing with the commonly used algorithm, has a better identification rate for identifying the healthy normal subject and heroin addict pulse waves.