Abstarct: Fingerprint is an important physiological feature used to identify individuals. Different methods have been proposed for fingerprint recognition. The present dissertation draws on fingerprint authentication methods and the results corresponding to these methods to develop new methods of fingerprint recognition. Gabor filter and segmentation were applied to fingerprint recognition in the first method. The background was removed from the input image, which was subsequently enhanced. Then, the center of finger was determined in the enhanced image. An area surrounding this center was identified and filtered in 8 directions using Gabor filter. The resulting 8 images were normalized. A new segmentation was applied to these normalized images and variance and mean values were determined as features for each segment. Feature extraction and recognition processes were lengthy since Gabor filter was used in this method. In the second method, Sobel Roberts Features (SRF) was used in feature extraction, which improved the time requirements compared to the first method. In this second method, the background was removed from the input image which was subsequently enhanced and prepared for recognition of the center. The enhanced image was then rotated around this center for 5 times from -45̊ to 45̊ with 22.5-degree steps. One area centered on the finger was identified in each rotation. These areas were normalized, and then, Sobel Roberts Features was used to extract feature vectors. The third method involved application of 2D Discrete Wavelet Transform (DWT), instead of Sobel Roberts Features, to the enhanced image for improvement in running time. First, DWT was used to obtain an approximated image to which Sobel Roberts Features was applied. The three methods mentioned above resulted in many unnecessary features which could be eliminated through binary particle swarm optimization (BPSO). BPSO was applied to the second method since in outperformed the other two methods. Euclidean distance and support vector machine (SVM) were used for the purpose of recognition by the three methods. Among the proposed systems, the second method had the highest efficiency, the third system had the best performance in terms of running time, and the fourth system had the smallest number of features with acceptable recognition rate. Databaxse from FVC2004 were used in this study.