人脸表情识别技术对微表情的研究分析

Research and Analysis of Micro-expressions by Facial Expression Recognition Technology

随着人们对情感智能的需求日益增大,情绪识别在复杂的社交中变得尤为重要。为减少微表情识别的误差,本文提出一种高效的微表情、微动作的识别算法。该算法使用fer2013/IMDB数据集、WiderFace数据集对YoLov5Face进行重训练,将YoLoV5Face检测到的人脸区域输入于Gabor小波变化、OpenCv实现对于外貌特征的提取,并利用AdaBoost机器学习算法实现表情的分类与识别。实验的结果表明,该算法具有较好的准确性和鲁棒性。

With the increasing demand for emotional intelligence,emotion recognition has become particularly important in complex social interactions.To reduce the error of micro expression recognition,this paper proposes an efficient algorithm for recognizing micro expressions and micro actions.This algorithm uses the fer2013/IMDB dataset The WiderFace dataset retrains YoLov5Face by inputting the detected facial regions into Gabor wavelet transform OpenCv implements the extraction of appearance features and utilizes the AdaBoost machine learning algorithm to classify and recognize facial expressions.The experimental results indicate that the algorithm has good accuracy and robustness.