基于异步多时域特征的动态手势识别方法

准确及时地手势识别在增强现实技术中具有重要的意义。针对表征复杂手势序列的时空特征，提出了一种基于异步多时域时空特征的手势识别方法。该方法通过轻量级三维卷积网络提取视频序列的不同时间步态的短期时空特征，通过改进的卷积长短期记忆网络学习长期时空特征，将不同步态的时空特征融合为异步多时域特征，以此来对手势进行分类识别。通过与其他主流方法进行比较，实验结果证明了提出的方法具有较高的动态手势识别率。

Dynamic Gesture Recognition Method Based on Asynchronous Multi-Time Domain Features

Accurate and timely gesture recognition is of great significance in augmented reality technology. Aiming at the spatio-temporal features that characterize complex gesture sequences, a gesture recognition method based on asynchronous multi-time domain features is proposed. It extracts the short-term spatiotemporal features of the video sequences with different temporal resolution through a lightweight three-dimensional convolutional network, and learns long-term spatio-temporal features through an improved convolutional long short-term memory network. The temporal and spatial features of different steps are fused into asynchronous multi-time domain features to classify and recognize gestures. Compared with other state-of-the-art approaches, the experimental results show that the proposed method has a high dynamic gesture recognition accuracy.