The high spectral resolution of hyper-spectral data provides the ability for diagnostic identification of various materials. The purpose of spectral feature extraction is to extract substantial information from the original data input and filter out redundant information,then classification and identification can be processed. The author introduced a method based on wavelet decomposition for spectral feature extraction. Appropriate wavelet is applied to the pre-analyzing signals and make it decomposed,then the high-frequency component contains much important feature information. We can detect every absorption strips and their left and right boundaries according to the original signal by finding the local modulus maxima of high-frequency weight. The method is terser and more effective than other traditional methods. It proved that wavelet analysis is a more perfect,method of spectral feature extraction.