Design and Research of Facial Expression Recognition System Based on Key Point Extraction

• Yan Qu,
• Yan Liu,

Vol. 19, No. 1, January 31, 2025
10.3837/tiis.2025.01.004, Download Paper (Free):

• feature extraction
• Image matching
• Spatio-temporal map convolutional networks
• Adaptive mechanisms
• Peak frame
• data fusion

Abstract

Currently, facial recognition is very common in the use of libraries, such as self-service borrowing systems and access gate systems. The face recognition system, which extracts facial key points, has become an integral part of people's daily lives and work. To enhance facial recognition accuracy, researchers have combined spatio-temporal graph convolutional network models with temporal feature information to extract facial expression features and recognize dynamic expressions through facial expression sequences. Additionally, they have introduced an adaptive attention mechanism and automatic adjustment of attention distribution for peak frame images, resulting in the acquisition of more comprehensive image information. The results indicated that the accuracy of the model increases by an average of 1.595% with the introduction of the adaptive module, resulting in a final recognition accuracy of 97.26%. Compared to the independent spatio-temporal graph convolutional network model, the accuracy was increased by an average of 7.65%. In conclusion, the proposed adaptive spatio-temporal graph convolutional network model based on peak frame image optimization has better facial expression recognition performance. This improved technology helps to improve the management efficiency of the library.

Statistics

Cite this article