A Novel and Efficient Feature Based Authentication System Using Multimodal Biometric Traits

Abstract

ABSTRACT newline newlineEvery activity in day-to-day life is required the need of mechanized automation for ensuring the security. The biometrics security system provides the automatic recognition of human by overcoming the traditional recognition methods like Password, Personal Identification Number (PIN) and Identity (ID) cards. Manual placement of card or entering of PIN / Password takes minimum 5 seconds to capture information where as biometric traits takes around 2 seconds and processing speed is between 0.6 to 0.8 seconds. newlineSince from the inception of digital computer the idea of recognizing humans with their specific physiological and behavioral traits has been a challenging task even after decades of research and deployments, the fields of biometric remains fresh even though new technologies are developed and old technologies are improved. Use of multimodal biometric for secured identification of an individual can reduce the false acceptance rate and false rejection rate parameters effectively, and also which provides more contention against imposter task due their uniqueness and universality. newline newlineMost biometrics are unimodal in nature, which rely on single source of information, but these systems currently suffer from noisy data, spoofing attacks, data quality and sometimes unacceptable error rates. These drawbacks can be overcome by setting up the multi-modal biometric systems newline newline newlineconsisting of two or more biometric modalities in a single identification system to improve the recognition accuracy and efficiency. newline newlineMulti modal biometrics is one of the most active and wide research area in many applications, it has gained a significant position with the interest of security. Existing multimodal biometric with the combination of common traits like finger print, face, palm, iris and voice provides accuracy ranging from 86% to 96%. But provides scope for improving efficiency and accuracy by employing enhanced techniques. newline newlineBiometric systems finds their application in various day to day activities starting from identifying individuals, time and attendance management in health, education, bank and business organization. System can also be deployed in military and defense related operations. newline newlineIn the proposed work we have designed and implemented a multimodal biometric system for face and voice modalities, where in which face recognition is carried out using Dual Tree Complex Wavelet Transform (DTCWT) integrated with predominant Quick Fourier Transform (QFT) technique results with 98.23% and voice recognition is carried using most powerful Mel Frequency Cepstral Coefficients (MFCC) combined with Relative Spectral Transform (RASTA) technique, which results with newline98.60% on L-Specek and Voxceleb data set respectively. newline newlineDTCWT is an enhancement technique over Discrete Wavelet Tree (DWT) with small additional properties and changes. It is an effective method for implementing an analytical wavelet transform, introduced by Kingsbury. DTCWT with generating complex coefficients introduces limited redundancy and allows the transform to provide shift invariance and directional selectivity of filters. newline newlineQFT is a straight forward Discrete Fourier Transform (DFT) which uses all possible symmetries of the DFT capable of considering shorter lengths of data and better suited for real data calculations with pruning. newline newlineMFCC method is used for voice authentication which is the most widely used and standardized technique for feature extraction in voice data. Which helps in reducing the frequency information of the input speech signal into coefficients, it is a very fast, reliable and easy computation method. RASTA is a special band-pass filter was added to smooth out short-term noise variations and to remove any constant offset in the speech channel which further increases the result. newline newlineThe discussion of results of face and voice data set over various databases are done with proposed face and voice recognition system. The results of experimentation are compared with existing methods and analysis proved that the proposed system placed in better position. newline newline