Study On Fingerprint Feature Extraction Based On Analytical Signals

With the widespread application of fingerprint identification system in various fields,the demand of automatic fingerprint identification system performance is also raised.Fingerprint recognition system is a typical pattern recognition system,including fingerprint acquisition,preprocessing,feature extraction and feature matching modules.Feature extraction module has a bearing on the role as the fingerprint identification system intermediate links.At present,the fingerprint feature extraction algorithm needs to observe the shape of the fingerprint image,understand the representation of the fingerprint feature in the image,and select the appropriate algorithm according to the morphological representation of the fingerprint image to operate the fingerprint image and obtain the feature points.The above is the commonly used feature extraction method at this stage.Is there a way to extract the features without the need to observe the morphological structure of the ridge?This paper proposes a study on fingerprint feature extraction method based on analytical signals,which is different from the traditional idea of image processing.This algorithm considers the fingerprint image as a fingerprint signal.Considering that the stripe pattern can be represented by a mathematical model,the fingerprint image is a special case of a stripe pattern.It can also be expressed as a twodimensional mathematical model.According to the established mathematical model,the fingerprint feature can be analyzed from the fingerprint signal.First of all,a twodimensional fingerprint mathematical model is established based on the stripe pattern.And the fingerprint image is regarded as a two-dimensional adjustable FM-AM signal.The commonly used fingerprint features are analyzed from the perspective of mathematical function.Secondly,the model contains irrelevant variables which are noise and compensation value.The two-dimensional fingerprint mathematical model is established,which is eliminated by Gaussian low-pass filtering,mean filtering and background segmentation.Thirdly,the fingerprint signal is further processed,and various features of the fingerprint are extracted according to the mathematical expression form of the fingerprint signal,including phase field,orientation field,minutiae,singularity point and ridge frequency.Finally,the international fingerprint identification game fingerprint database is used to set up an experimental platform.The features extracted by the algorithm and the features extracted by the classical algorithm are applied to the automatic fingerprint identification system and the experimental results are analyzed.The experimental results show that the proposed algorithm can ignore the changes of the coordinates of the feature points caused by finger pressure,finger wetness and finger position,etc.And the proposed algorithm can extract the features that can be resolved in the fingerprint signal,eliminate the need to delete the pseudofeature points generated due to the disconnection and reduce the errors caused by unnecessary links.This algorithm can ensure system identification accuracy under the premise of reducing system uptime and improve the performance of the identification system.