Unfamiliar Dynamic Hand Gestures Recognition Based on Zero-Shot Learning
文献类型：会议论文
| 作者 | Jinting Wu1,2 ; Kang Li1,2; Xiaoguang Zhao2; Min Tan2
|
| 出版日期 | 2018-12 |
| 会议日期 | 2018-12 |
| 会议地点 | Siem Reap, Cambodia |
| 关键词 | Dynamic Hand Gesture Recognition Bidirectional Long-Short-Term Memory (BLSTM) Zero-Shot Learning (ZSL) Semantic Autoencoder (SAE) Leap Motion Controller (LMC) |
| 英文摘要 | Most existing robots can recognize trained hand gestures to interpret user's intent, while untrained dynamic hand gestures are hard to be understood correctly. This paper presents a dynamic hand gesture recognition approach based on Zero-Shot Learning (ZSL), which can recognize untrained hand gestures and predict user's intention. To this end, we utilize a Bidirectional Long-Short-Term Memory (BLSTM) network to extract hand gesture feature from skeletal joint data collected by Leap Motion Controller (LMC). Specififically, this data is used to construct a novel dynamic hand gesture dataset for human-robot interaction application. Twenty common hand gestures are included and fififteen concrete semantic attributes are condensed. Based on these features and semantic attributes, a Semantic Autoencoder (SAE) is employed to learn a mapping from feature space to semantic space. By matching the most similar semantic information, the unfamiliar hand gestures are recognized as correct as possible. Experimental results on our dataset indicate that the proposed approach can effffectively identify unfamiliar hand gestures. |
| 源URL | [http://ir.ia.ac.cn/handle/173211/49704]  |
| 专题 | 自动化研究所_复杂系统管理与控制国家重点实验室_先进机器人控制团队 |
| 通讯作者 | Jinting Wu |
| 作者单位 | 1.University of Chinese Academy of Sciences 2.Institute of Automation, Chinese Academy of Sciences |
| 推荐引用方式 GB/T 7714 | Jinting Wu,Kang Li,Xiaoguang Zhao,et al. Unfamiliar Dynamic Hand Gestures Recognition Based on Zero-Shot Learning[C]. 见:. Siem Reap, Cambodia. 2018-12. |
入库方式： OAI收割
来源：自动化研究所
浏览0
下载0
收藏0
其他版本
除非特别说明，本系统中所有内容都受版权保护，并保留所有权利。