Haiyan Ye, and Qilin Wu
Facial expression, posture, depth recognition, cyber-physical systems
Due to the limitations of traditional face recognition methods, such as a single posture, numerous model parameters, and the impact of environmental factors, such as lighting, feature extraction speed can be slow. To effectively address these issues, a new bi-modal emotion- based deep learning face recognition method has been proposed that combines both emotional and postural information. This approach significantly improves feature extraction speed and provides a more efficient and effective solution for face recognition. By extracting facial expression and pose features, a multi-pose face data set is established, and the dual-mode emotion depth recognition method of expression and posture is studied. The convolution neural network algorithm based on multi-task cascade is used to collect facial expression and advanced pose features for face tracking, determine the key points of face pose, extract the arc surface of loss function, and compare with the corresponding pose features in the database, finally obtain the result of facial expression and pose bimodal emotion depth recognition. Experimental results show that the method can significantly improve the accuracy of face recognition and the speed of feature extraction.
Important Links:
Go Back
