摘要
动态手势识别是手势交互的关键技术,针对动态手势数据的时序性和空间不确定性造成识别困难问题,提出一种基于卷积神经网络(CNN)和长短期记忆网络(LSTM)相结合的动态手势序列识别方法。实验采用数据手套采集动态手势数据,对定义的8种动态手势进行测试,平均识别率达到了92.5%。实验表明,与单纯使用LSTM模型或CNN模型对比,所提模型识别率较高,在虚拟现实界面交互任务中用户体验更好。
Dynamic gesture recognition is the key technology of gesture interaction. Aimed at the difficulty of recognition caused by the timing and spatial uncertainty of dynamic gesture data, a dynamic gesture sequence recognition method based on convolutional neural network(CNN) and long-term short-term memory network(LSTM) is proposed. In the experiment, data gloves were used to collect dynamic gesture data, and the eight defined dynamic gestures were tested. The average recognition rate reached 92.5%. Experiments show that compared with the LSTM model or CNN model alone, the proposed model has a higher recognition rate and a better user experience in virtual reality interface interaction tasks.
作者
谷学静
周自朋
郭宇承
李晓刚
Gu Xuejing;Zhou Zipeng;Guo Yucheng;Li Xiaogang(School of Electrical Engineering,North China University of Technology,Tangshan 063210,Hebei,China;Virtual Simulation Experimental Teaching Center of Metallurgical Engineering,Tangshan 063000,Hebei,China;Information Automation Department,Tangshan Iron and Steel Group,Tangshan 063000,Hebei,China)
出处
《计算机应用与软件》
北大核心
2021年第11期205-209,共5页
Computer Applications and Software
基金
河北省自然科学基金高端钢铁冶金联合研究基金专项项目(F2017209120)。
关键词
虚拟现实
人机交互
动态手势识别
卷积神经网络
长短期记忆网络
混合模型
Virtual reality
Human-computer interaction
Dynamic gesture recognition
Convolutional neural network
Long and short-term memory network
Hybrid model
