Palmprint recognition based on Mel frequency Cepstral coefficients feature extraction

Palmprint identification is a measurement of palmprint features for recognizing the identity of a user. Palmprint is universal, easy to capture and it does not change much across time. This paper presents an application of Mel frequency Cepstral coefficients (MFCCs) for identification of palmprint. Palmprint feature extraction is based on transforming the palmprint image into one dimensional (1-D) signal and then extracting MFCCs from this signal. Wavelet transform (DWT) of the 1-D palmprint signals are used for extracting additional features to help in the recognition process. The features from MFCCs of this DWT vector are added to the MFCCs feature vector, generated from the original palmprint signal, to form a large feature vector that can be used for palmprint identification. Feature matching is performed in this research using feed forward back propagation error neural network. Experimental results show that the proposed method is robust in the presence of noise.