基于深度学习模型的自发学习表情识别方法研究

课堂教学环境中,面部表情自动识别是获取学习者情绪状态的重要方式。针对传统方法提取特征不全面、融合特征维度较高等问题,提出一种融合局部与全局特征的学习表情自动识别方法。该方法提取并融合表情图像的局部几何特征、KPCA降维的CLBP全局浅层纹理特征和CNN全局深度网络特征。此外,还构建一个全新的自发学习表情数据库,将课堂学习中的情绪分为困惑、快乐、疲倦、惊讶和中性等5种类型,用于CNN模型的训练。对比实验表明,该方法的识别正确率在CK+库、中国情绪图片系统和自发学习表情数据库中分别达到96.3%、86.7%和95.6%,高于传统的面部表情识别方法。该方法能够有效获取课堂中学生情绪变化,帮助教师准确全面地掌握班级学生的整体情况,促进课堂教学质量的提高。

SPONTANEOUS LEARNING FACIAL EXPRESSION RECOGNITION BASED ON DEEP LEARNING

Automatic facial expressions recognition is an important way to acquire the emotional state of the students in the classroom. The traditional methods often suffer from the problems of incomplete feature extraction and high dimension of fusion features. In this paper, we proposed an automatic expression recognition algorithm by extracting the local and global features of the facial expression images. The proposed algorithm extracted and fused the multi-features in facial expression images, such as the local geometric features, global shallow texture features of KPCA dimension reduced CLBP and CNN global depth network features. Moreover, we constructed a new spontaneous learning expression database. In the database, all the emotions images were classified into five types: confusion, happiness, fatigue, surprise and neutral. The developed database was utilized for the training of CNN models. Experimental results show that the recognition rates of this method in CK+ database, Chinese emotion image system and autonomous learning expression database are 96.3%, 86.7% and 95.6% respectively. It is higher than the traditional facial expression recognition methods. The proposed method can recognize the emotional changes of students in the classroom effectively. It is helpful for teachers to grasp the overall situation of classroom students accurately and comprehensively, and promote the improvement of classroom teaching quality.