摘要
针对复杂背景下的手势识别问题,提出了一种基于改进AlexNet的手势识别方案.根据手势识别问题的特点,方案对训练集中的图片和待识别的图片使用Sobel算子进行边缘提取,既保留了手势的轮廓,又屏蔽了脸部、手臂等裸露的皮肤对识别结果的影响;同时通过对AlexNet网络的结构优化和超参数的优化选择提高了模型的性能.实验结果显示,方案在保证实时性的前提下,能够达到约93%的识别准确率.
To solve the problem of hand gesture recognition in complicated situation,a method based on improved AlexNet is proposed.Sobel operator is used to extract the edge of the image both for image in the training set and that to be recognized,which can keep the outline of the gesture while discard some useless information.At the same time,the performance of the model is improved by optimizing the structure of AlexNet and the selected super parameters.The experiment results show that the method can achieve 94%recognition accuracy under the premise of real-time.
作者
郭书杰
GUO Shujie(School of Intelligence&Electronic Engineering,Dalian Neusoft University of Information,Dalian 116023,China)
出处
《大连交通大学学报》
CAS
2020年第6期95-99,共5页
Journal of Dalian Jiaotong University
