A New Effective Wearable Hand Gesture Recognition Algorithm With 3-axis Accelerometer

In this paper, a new simple and effective hand gesture recognition algorithm with 3-axis accelerometer sensor is proposed. It is different from much of the previous work that in our algorithm the preprocessed 3-axis waveforms from accelerometer are directly used as the input features and a fast dynamic time warping(FastDTW) is used as the recognizer. Our proposed algorithm is device-independent in that it can be very easily implemented and portable on any other smart wearable device equipped with 3-axis accelerometer. The experimental results show that the proposed algorithm can perform 99% correct rate with 5 training samples for each gesture in our dataset, which consists of 160 test samples from 16 different gestures. Besides, our algorithm can be 8 times faster than standard dynamic time warping(DTW) algorithm without performance decline.