Fingerprint Identification System Based On Combining Wavelet Packet Transformation With PNN Neural Network

In fingerprint identification, voice recognition, retina recognition and other biometric authentication (identification) technology, because of fingerprint identification has many advantages--uniqueness (each person's fingerprint is unique), invariance (fingerprint has strong relative stability), practicability (fingerprint sample is easy to acquire), and safety (each person's ten fingers have different fingerprint, so they can constitute multiple password), etc, fingerprint identification has applied widely.While fingerprint is only a small part of human skin, because of the rather large identifying quantity, the noise of fingerprint images, the non-linear elastic deformation of the skin, the position offset of the fingerprint input and other factors, fingerprint identification process doesn't simply contrast the equality or inequality of the fingerprint data, but needs complex image processing, fuzzy matching and identification algorithm, so it is very difficult to study. Adding to the fact that many research documents are not open for business interests, although many shaping products are created, ideal AFIS (Automated Fingerprint Identification System) is still a very difficult research work, particularly there are still some problems in the fingerprint image preprocessing, automatic identification, etc, needing to further explorate and study.AFIS consists of three main steps: image preprocessing, image feature extraction, image feature matching. Studying every aspect of fingerprint identification system, this paper brings up a kind of fingerprint identification algorithm based on combining wavelet packet transformation with PNN neural network. Through simulation experiments, some conclusions can be drawn as follows:1. Through this paper, we can know that using the algorithm of based on combining median filtering with wavelet transformation can effectively eliminate the interference of random noise to image caused by fingerprint image acquisition process, and draw a conclusion that when noise coefficient is less than 0.25, the denoising ability of wavelet packet transformation is better than wavelet transformation, which proves the ability of wavelet packet transformation to analyze image details.2. Adding the assessment of image quality to image preprocessing can effectively remove the influences of dry, wet and incomplete fingerprint to the fingerprint identification rate, and also the quality assessment algorithm based on binarization is simple to realize, has fast processing speed, and can meet real-time requirements. 3. Normalizing the size and gray of images can remove images'redundant information, and make fingerprint images become gray uniform, which give prominence to the main messages of images and effectively reduce the difficulties of fingerprint feature extraction caused by image redundancy and uneven gray.4. After decomposing the fingerprint image with wavelet packet and optimizing the Quad tree of decomposed image, the optimal wavelet tree is obtained. By calculating leaf nodes'energy value of the optimal tree, the fingerprint feature values are obtained, which are used to represent fingerprints. This method can obtain few & accurate feature point and greatly increase processing speed of the system.5. PNN neural network combines the advantages of the RBF neural network and competitive neural network. The simulation results show it has strong ability both in pattern classification and pattern recognition, and need very short time to build and train a network.On the base of deeply analyzing the characteristics of fingerprint image, fingerprint image preprocessing, feature extraction and pattern recognition, this paper studies the component, principle and identification method of the fingerprint identification system based on combining the wavelet packet transformation and neural network, and analyzes the system's strengths and weaknesses, which provides the basis and method to establish practical fingerprint identification system.