IEEE Trans Biomed Circuits Syst. 2019 Dec;13(6):1425-1436. doi: 10.1109/TBCAS.2019.2940030. Epub 2019 Sep 9.
This paper proposes a novel framework to process pressure signals for real-time and robust gesture recognition, which includes an innovative segmentation scheme, a gesture recognition scheme and a pressure-parameter adaptive updating strategy. A prototype system, including a wearable gesture sensing device with four pressure sensors and the corresponding algorithmic framework, is developed to realize real-time gesture-based interaction. With the device worn on the wrist, the user can interact with the computer using 8 predefined gestures. Experimental results show that the delay of gesture recognition is about 100 ms, with the average accuracy of 95.28% in the experienced-user test and 86.20% in the inexperienced-user test. Finally, the system is evaluated by a mouse-controlling interaction task and performs well. Both experienced and inexperienced people can easily and quickly complete interactive tasks. These results demonstrate that a pressure-sensor based wristband can be used to classify hand gestures well and to control the mouse interaction. This approach provides an interactive way to replace the mouse for decreasing the risk of the carpal tunnel syndrome (CTS).
本文提出了一种新颖的框架来处理压力信号，以实现实时和鲁棒的手势识别，该框架包括创新的分段方案、手势识别方案和压力参数自适应更新策略。开发了一个包括带有四个压力传感器的可穿戴手势感应设备和相应的算法框架的原型系统，以实现基于实时手势的交互。用户将设备戴在手腕上，可以使用 8 个预定义的手势与计算机进行交互。实验结果表明，手势识别的延迟约为 100ms，在有经验的用户测试中的平均准确率为 95.28%，在无经验的用户测试中的准确率为 86.20%。最后，通过鼠标控制交互任务对系统进行了评估，表现良好。有经验和无经验的用户都可以轻松、快速地完成交互任务。这些结果表明，基于压力传感器的腕带可以很好地对手势进行分类，并控制鼠标交互。这种方法提供了一种替代鼠标的交互方式，以降低腕管综合征（CTS）的风险。
