基于深度学习的图像人脸识别方法研究

Research on image face recognition method based on deep learning

使用计算机进行面部表情识别是当前人脸面部表情识别的热点,在深度学习技术的基础上,应用级联分类器对面部进行整体检测和分区定位后,提出并使用了一种基于自注意力机制的深度卷积神经网络,模型采用Mini-Xception为基本网络融合了注意力机制,再通过训练卷积神经网络构建表情分类模型,最后实现较为快速准确的表情识别。文中采用几种方法进行实验对比,并对最终的实验结果加以分析。结果表明,在相同的参数设置下提出的方法能明显提高分类性能、识别的精准度以及面部表情变化检测的实时速度。

Facial expression recognition using computer is the focus of facial expression recognition.In order to put this direction into production and life more effectively.In this paper,on the basis of deep learning technology based on cascade classifier for face detection and partition the whole positioning,based on the attention mechanism is put forward and used a,the depth of the convolution neural network model of the Mini-Xception as the basic network is a blend of attention mechanism,and convolution model of neural network to construct expression classification through training,Finally,fast and accurate facial expression recognition is achieved.In this paper,a variety of methods are used for experimental comparison,and the final experimental results are analyzed.The results show that,under the same parameter setting,the proposed method can significantly improve the classification performance,recognition accuracy and real-time speed of facial expression change detection.