Research on Multimodal Biometric Computer-Assisted Identification System Based on Deep Learning

This paper puts forward a multimodal biometric recognition system founded on deep learning. Through integrating various biometric data like fingerprints, faces, and irises, the precision and sturdiness of recognition are observably enhanced. The method of combining CNN and LSTM is used to fully explore the feature interaction relationship of different modal data. In the process of model training, this paper introduces an adaptive weighted loss function to dynamically adjust the influence weights of different modal features, thereby enhancing the stability and anti-interference ability of recognition. This system is tested using a public data set containing many multimodal biometric features and compared with traditional recognition methods. The results show that the system has achieved significant improvements in both recognition accuracy and response speed. Specifically, the system achieved an accuracy rate of 98.7% under the fusion of different modalities, which is about 12.5% higher than that of a single modality. In addition, this system also shows higher robustness when processing noisy data, and the recognition error rate is reduced to 2.3%.