A technique for classification of galaxy spectra is proposed.At first

a four-level wavelet decomposition of the original spectrum is performed

and the wavelet coefficient at the fourth level

which mainly includes the information of spectral lines

is chosen as the wavelet feature of the spectrum.Secondly

principal components analysis is used to compress the wavelet feature and to get the recognition feature of the spectrum.Finally

Fisher linear discriminant analysis is employed for classification.This approach can recognize the galaxy spectrum whose flux is uncalibrated and redshift is unknown.Comparisons with several other classification techniques were made by experiments.Experiment results show the proposed method is robust and superior to other methods under the condition that flux is uncalibrated.