Micro-expression sequence recognition method based on double-flow Vision Transform

• Main Authors: ZHU HAONAN, ZHOU WENHUI, ZHANG HUA, DOU XIAOLONG, DAI GUOJUN
• Format: Patent
• Language: Chinese; English
• Published: 29.09.2023
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; PHYSICS

The invention discloses a micro expression sequence identification method based on a double flow VisionTransform. The method comprises the following steps: firstly, acquiring a plurality of micro-expression images in continuous time, and preprocessing the micro-expression sequence images to obtain preprocessed micro-expression sequence images and optical flow images; obtaining fusion features of each optical flow image and the corresponding micro-expression image through a TS-ViT module; learning spatio-temporal features of continuous frames through a time attention mechanism module, and performing feature fusion to obtain a new feature vector sequence; and inputting the new feature vector sequence into the Bi-LSTM to learn the forward and backward dependency relationship of the micro-expression sequence in the time dimension, and finally outputting a micro-expression recognition result. According to the method, the problem that network training is under-fitted due to the fact that the data size of a micro-ex