基于改进YOLOV3与贝叶斯分类器的手势识别方法研究

Research Approach of Hand Gesture Recognition Based on Improved YOLOV3 Network and Bayes Classifier

手势识别研究与人机交互和谐发展具有密不可分的联系,因此具有重要研究意义.针对传统手势检测算法空间不变性较弱,手势识别效率较低等问题,本文提出基于改进YOLOV3网络与贝叶斯分类器相结合的手势识别深度学习模型.首先采用空间变换网络对YOLOV3网络进行改进,处理手势信息,提取关键性手势特征,解决了数据易受影响问题并且增强了网络不变性;然后将网络提取出的特征进行降维操作,减少冗余信息;再通过贝叶斯分类器进行分类,提高了分类准确率;最后在标准数据集和自制数据集上进行检测测试,表明本文方法能够提高手势的识别精度,验证了算法的有效性.

The study of gesture recognition is closely related to the harmonious development of human-computer interaction,so it has important research significance.In view of the weak spatial invariability and low gesture recognition efficiency of traditional gesture detection algorithms,this paper proposes a gesture recognition model based on the improved YOLOV3 network combined with Bayesian classifier.Firstly,the spatial transformation network was used to improve the YOLOV3 network,process gesture information,extract key gesture features,and solve the problem of data vulnerability and enhance network invariability.Then the features extracted from the network are reduced to reduce the redundant information.The classification accuracy is improved by using bayesian classifier.Finally,the algorithm is tested on standard data set and homemade data set.The method in this paper improves the recognition accuracy of gestures and proves the effectiveness of the algorithm.