This paper presents a pattern recognition approach which is developed for biometric identification using individual's knuckle prints. In this approach, initially palm images are segmented with active appearance models and regions of interest (knuckle prints) are extracted with using analytical processing. Afterwards, the patterns of knuckle prints are extracted by combining these regions. First, discrete wavelet transform are applied to transform to the spectral domain for feature extraction, then by using nonlinear Kernel Fisher Discriminant method most discriminative features are obtained. Weighted Euclidean distance based nearest neighbor method is utilized for classification. Finally, the proposed method is tested on 1614 hand images which belong 132 different persons. Obtained results (%97 accuracy rate for 132 persons) demonstrate proposed method's success, they are promising for the future.