Biological tissues are identified based on unique features of their Raman spectra. Raman signal data of biological tissues is acquired by a self-designed Raman probe, and preprocessed to rectify the baseline through filtering noises and stray light. The principal component analysis method is used to extract the critical Raman signal features of biological tissues, and then a back propagation (BP) neural network algorithm is used to classify the biological tissues by using these features. Automatic classification is implemented with the Raman spectrum data from the animal tissue phantoms. Experimental results show that the BP neural network is efficient to identify different animal tissues, and the accuracy rate reaches nearly 95%.