基于ResNet和HOA模型的面部表情识别方法研究

Research on Facial Expression Recognition Method Based on ResNet and HOA Model

针对面部表情识别率低的问题,提出了一种ResNet结合HOA的面部表情识别方法。首先,利用复杂的高阶统计信息,对残差网络提取的特征进行建模,进而提高表情识别的准确率;其次,通过对ResNet模型进行改造,使其适用于灰度图像;最后,在CK+面部表情数据集和Fer2013面部表情数据集上进行实验,验证设计方法的有效性。结果显示,所提出方法在两个数据集上的识别准确率可分别高达97.50%和94.40%。通过对比实验表明,相比其他几种较新的面部表情识别方法,提出的方法可以获得更高的识别率。

To solve the problem of low recognition rate of facial expression,a method of facial expression recognition based on ResNet and HOA is proposed.First of all,the complex higher-order statistical information is used to model the features extracted from the residual network,so as to improve the accuracy of expression recognition;Secondly,the ResNet model is modified to adapt to grayscale images;Finally,experiments were conducted on CK+facial expression dataset and Fer2013 facial expression dataset to verify the effectiveness of the design method.The results show that the recognition accuracy of the proposed method on the two datasets can reach 97.50%and 94.40%respectively.The comparative experiments show that the proposed method can achieve higher recognition rate than other several relatively new facial expression recognition methods.