基于加速度传感器的连续动态手势识别

Continuous dynamic gesture recognition based on acceleration sensor

针对个体手势动作信号的差异性和不稳定性,提出了一种基于加速度传感器的连续动态手势识别方法。通过MEMS加速度传感器采集手势动作信号,并结合手势信号的动作特征,对单个手势的有效数据进行自动定位截取,经预处理和特征提取后,构建隐马尔可夫模型(HMM)以实现对特定手势的实时识别。通过设计实现了一种可穿戴手势信号采集硬件原型系统,对10类手势的1000个手势数据进行识别对比实验,统计结果表明:该方法可以对连续手势进行实时有效的识别。

Aiming at differences and instability of individual gestures signal, a continuous dynamic gesture recognition method based on acceleration sensor is proposed. The method utilizes MEMS acceleration sensor to capture gesture acceleration signal, according to the features of gesture signal, valid data of single gesture is automatically located and intercepted, after pre-processing and features extraction hidden Markov models （ HMM ） to real-time identify specific gesture. A wearable gesture signal acquisition hardware prototype system is realized through design, 1000 data of 10 types of gesture is selected to conducte the identification and comparative experiment, statistical results show that the method can identify continuous gesture in real-time and effectively.