Enhanced Multimodal Biometric Fusion with DWT, LSTM, and Attention Mechanism for Face and Iris Recognition

This paper introduces a new multimodal biometric system integrating facial and iris recognition using discrete wavelet transform (DWT) and long short-term memory (LSTM) neural networks enhanced with an attention mechanism. DWT is effective in extracting global and local features, and these extracted traits are addressed as instances of serial data to enable LSTMs to pick temporal connections in and provide context-specific continuations with continual regularity; other than these factors, improvement of the features has a major impact because attention emphasizes exactly on where more contextual contribution would occur with these elements to allow further contextual insights during the output phase. Experimental results show commendable gains in performance as compared with conventional systems, and achieves identification of up to $\mathbf{9 9 . 9 4 \%}$. It can be utilized in security environment where accurate identification with biometrics and proper authentication is highly needed. Framework: Features acquired from the DWT are integrated inside an LSTM network. Thus, it provides guidelines for both the structured feeding schedule and the organizing of time data, thus better facilitating biometric recognition. Though some progress has been made in optimizing biometric identification for individuals with varied features, there is still much to be done.