人脸表情识别是计算机视觉领域中人脸识别的一个重要分支。由于人脸表情多样性，头部姿态变化以及表情主 体所处环境等诸多因素的影响，给人脸表情识别的工作带来了很大的挑战。针对采用传统卷积神经网络，由于其模型参数数 量多，且比传统机器学习算法的人脸表情识别精度的提高有限，给出了一种基于深度可分离卷积结构的改进卷积神经网络模 型。基于该模型对Fer2013灰度表情识别数据集进行实验，结果表明，在保证了68.31% 的较高准确率情况下，与传统卷积神经 网络相比，模型的网络结构得到了优化，模型参数数量大大减少，且模型参数的利用效率较高。

Facial expression recognition is an important branch of facial recognition in the field of computer vision. Influenced by many factors including the diversity of facial expressions, changes in head posture and the environment, it is a big challenge to the work of facial expression recognition. Focused on traditional convolutional neural network, owning to its large number of parame- ters and the limited improvement in facial expression recognition accuracy compared with applying traditional machine learning al- gorithms, we proposes an improved convolutional neural network model based on the depthwise separable convolution structure. We make the experiment on the Fer2013 gray expression recognition dataset based on our model, the result shows that the network struc- ture of the model is optimized, the number of model parameters is greatly reduced, and the utilization efficiency of model parame- ters is high, while ensuring a high accuracy rate of 68.31% compared to the traditional convolutional neural network.