Fingerprint Liveness Detection with Machine Learning Classifiers using Feature Level Fusion of Spatial and Transform Domain Features

Ease of accessibility with unique person identification has made biometric more popular. One of the most extensively studied and used biometric identification is fingerprint. These days false imitation of fingerprints created using wood glue, gelatin, cadaver are posing challenges for biometric identification systems dependent on fingerprint trait.. Hence the research area of fingerprint liveness detection has become very primal and important. The Machine learning classifiers are facilitating in the fingerprint liveness detection with the aid of fusion of extracted of fingerprints of indeed features. The fingerprint ridge frequencies in spatial domain are taken in vertical, horizontal, backward diagonal and forward diagonal directions as the features. Due to the energy compaction characteristics of orthogonal transforms are enabling the suitable extraction of features. The fractional coefficients of Cosine transformed fingerprint images are considered as feature vectors. For faster and more accurate fingerprint liveness detection,. The paper proposes innovative fingerprint liveness detection technique with fusion of fingerprint ridge directional frequencies and the fractional coefficients of Cosine transformed fingerprints with nine distinct machine learning algorithms . The experimentation is conducted with fusion of four directional ridge frequencies and 0.024 percent of fractional coefficients of Cosine transformed fingerprint images taken to constitute the feature vectors with nine diverse machine learning algorithms. The proposed fusion based techniques is explored with method and tested on two of the benchmark datasets alias ‘ATVS’ and ‘FVC 2000’. For performances of assessment of the variants of proposed fusion based fingerprint liveness detection techniques , the performance measures used are percentage classification accuracy, percentage precision and percentage recall. The feature level fusion of directional fingerprint ridge frequency and 0.024% of fractional coefficients of Cosine transformed fingerprint image has shown the improvement in percentage classification accuracy of fingerprint liveness detection.