基于深度学习与传统机器学习的人脸表情识别综述

Facial expression recognition based on deep learning and traditional machine learning

现有的人脸表情识别技术基本局限于传统的机器学习算法,在光照强弱、有遮挡物、姿态变换等情况下,传统的机器学习算法鲁棒性差,难以运用到实际生活中。随着计算机GPU等硬件条件的发展、大数据时代的到来,深度学习在计算机视觉领域备受关注。本文从图像预处理、特征提取、特征分类等方面介绍了传统机器学习算法及其优缺点;从DBN、CNN等主流算法、发展方向、常用开发框架介绍了深度学习算法。最后总结和展望了传统机器学习与深度学习在人脸表情识别上的发展问题与趋势以及后续研究方向。

The existing facial expression recognition technology is limited to the traditional machine learning algorithms basically. The traditional machine learning algorithms have a low robustness in the case of light intensity, obstruction and posture change, which causes some difficulties in the practical application. With the development of the computer hardware conditions such as GPU and the development of the big data, deep learning became the focus in the field of computer vision. This paper introduces the traditional machine learning algorithm, its merits and shortcomings from such aspects as image preprocessing, feature extraction and feature classification; describes the deep learning algorithm from such aspects as the mainstream algorithms including DBN and CNN, development direction and common development framework ; finally, summarizes and forecasts the development and trends of traditional machine learning and deep learning in facial expression recognition as well as the follow-up research direction.