A dual-modality human feature recognition method using matched layer fusion

• Published in: Scientific Bulletin. Series C, Electrical Engineering and Computer Science no. 2; p. 234
• Main Authors: Deng, Ke, Mo, Jiawei, Huang, Rongping
• Format: Journal Article
• Language: English
• Published: Bucharest University Polytechnica of Bucharest 01.01.2025
• Subjects: Biometrics; Contour matching; Data integration; Face recognition; Feature extraction; Feature recognition; Matching layers (electronics)
• ISSN: 2286-3540

The objective of this paper is to present a recognition method based on the fusion of two human features, namely gait and face, within a matching layer. Spacetemporal biometric features with differentiation in human contour maps are obtained by gait feature extraction network in order to resolve the issue that facial recognition technology is difficult to recognize the target subject with high accuracy under the condition of having interfering objects on the face or longer detection distance. A facial feature extraction network is employed to obtain fine-grained features of the face in order to enhance the immunity of the network to the conditions where the contour of the target subject is affected by interfering objects. Facial features are fused with gait features at the matching layer for information fusion in order to achieve complementarity between the two modal biometrics. The experimental results show that the method proposed in the paper has higher recognition accuracy compared to the gait or facial feature recognition methods in unimodal mode.