Multimodal Biometric System Using Dual Digital Watermarking

Abstract

Multimodal biometric system relies on multiple biometric identifiers for person identification that becomes more popular in authentication and identification. Due to the popularity of multimodal system, three factors (security, authentication and robustness) are considered for image authentication, face recognition and biometric identification. This paper presents multimodal biometric system by applying double digital watermarking scheme. The proposed dual watermarking is embedding blind semi-fragile and robust watermarks into the facial image. To embed blind semi-fragile digital watermarking, IWT (Integer Wavelet Transform) and Reversible Procedure is proposed and for robust digital watermarks, DCT (Discrete Contourlet Transform) and QIM (Quantization Index Modulation) is proposed. In this, we have used two biometrics including face and voice for biometric system. For watermark embedding, MFCC (Mel Frequency Cepstral Coefficients) voice features are extracted and embedding into facial image. Face selection operation is performed using ICP (Iterative Closest Point) algorithm that works based on learning weights.  Final decision making is performed using Deep Reinforcement Learning called “Double Deep-Q-Network”. Two corpuses such as TIMIT (voice) and ORL (Face) are used for system evaluation and performance testing. our proposed double watermarking scheme exhibits better performance in terms of accuracy, EER, PSNR, SSIM,