A Multimodal Identity Recognition Algorithm

Although biometrics technology has developed for many years, in practical applications, it still faces many difficulties and challenges. In non-ideal, uncontrolled, unconstrained scenarios, the recognition accuracy is greatly reduced. Although good recognition results have been obtained based on single-mode information, most algorithms are built based on constrained environment, and the data used are also taken from constrained environment. If the algorithms encounter unconstrained situations, it is not reliable to recognize only a single mode such as facial features, and the accuracy of face recognition [1] is even lower when there is facial occlusion or insufficient light. In this paper, a multi-modal [2] knowledge fusion algorithm is proposed, which fuses facial features, human features and voiceprint features, and comprehensively studies modal biometric information to improve the robustness and accuracy of the identity recognition algorithm.