Abstract:

This paper proposes a principal component analysis method combined with wavelet low-frequency band todeal with the face recognition with single training sample. To enhance the classification information of a single sampleimage , the method combines the original training image with its reconstructed image based on wavelet low-frequencyband , and performs the principal component analysis on the joined version of the training image. Experimentalresults on the ORL database show that , when each person has only one training sample , the proposed methodachieves 3. 6% higher recognition accuracy and uses 14. 8% fewer eigenfaces than the standard eigenface algorithm.

Key words: face recognition, wavelet transform, principal component analysis, eigenface