Multimodal Biometrics Classification Using Long Short Term Memory with Sigmoid Functions

In the evolving fields of biometric authentication and recognition, ongoing efforts are aimed at overcoming various challenges related to accuracy, robustness, computational complexity, and generalization. Notably, there are common challenges such as the need for image augmentation, dataset size impact, and architectural limitations in deep learning approaches. To address these issues, this paper leverages two diverse datasets, the CASIA dataset, and the UCI-HAR dataset, each offering unique biometric data sources. The proposed method employs Long Short Term Memory (LSTM) with a sigmoid neural network for classification, effectively capturing temporal patterns and enhancing biometric recognition. This approach has the potential to revolutionize biometric classification and authentication, offering enhanced accuracy, robustness, and applicability across various scenarios. The performance of this method is exemplary, achieving an accuracy rate of 99.80%. Furthermore, the proposed method is compared against established techniques, including Support Vector Machine (SVM), K-Nearest Neighbor (K-NN), and Deep Convolutional Neural Network, highlighting its superior performance. By effectively addressing the prevalent challenges in the field, it paves the way for more reliable and secure biometric recognition systems.