Authors: Nianhui Wang; Qingxue Li; Yuanhua Li
Addresses: College of Physical Education and Health, Linyi University, Linyi, 276000, China ' College of Physical Education and Health, Linyi University, Linyi, 276000, China ' College of Physical Education, Central South University of Forestry and Technology, Changsha 410004, China
Abstract: Aiming at the problems of high loss rate of expression details, poor comprehensiveness of recognition results and low recognition rate in traditional methods, an athlete's facial emotion recognition based on Kalman filter is proposed. Firstly, the uncertainty of athlete's facial emotion image is described according to the active learning algorithm. Then, with full consideration of the uncertainty factors, the feature block method is used to mosaic the image. Finally, according to the splicing results, a first-order motion model is established by Kalman filter to track and calibrate the image target points to complete the facial emotion recognition. The results show that the expression detail loss rate of this method is low, the comprehensiveness coefficient of recognition results is high, and the emotion recognition rate is always higher than 90%, indicating that the recognition effect of this method is better.
Keywords: Kalman filter; emotion recognition; active learning algorithm; first-order motion model; feature block; image mosaic; uncertainty description.
International Journal of Reasoning-based Intelligent Systems, 2022 Vol.14 No.4, pp.221 - 226
Received: 27 Oct 2021
Accepted: 13 Apr 2022
Published online: 31 Oct 2022 *