Xu Zhang et al., “An Adaptive Feature Extractor for Gesture SEMG Recognition,” in Medical Biometrics, ed. David Zhang, vol. 4901 (Berlin, Heidelberg: Springer Berlin Heidelberg, 2007), 83-90

1. Surface Electromyographic sEMG sensor can help in virtual input devices productions. It use a technique for evaluating and recording the electrical activity produced by skeletal muscles.[1]

2. One of the sEMG problems is the reusability. Most of the current proposed sEMG pattern identification methods only work on signals from the same use sessions. However, even on the same subjects, the sEMG measured in one day is relatively different from that in another day. The differences of sEMG between different use sessions are mainly caused by slight misplacement of the reinstalled sensors and individual changes of subjects.

3. The performance of hand gestures recognition may be improved consequently with higher pattern recognition rate and pre-training

4. The effect of training days on the average recognition rates of six kinds of hand motions for each subject is; the performance of hand gestures recognition may be improved consequently with higher pattern recognition rate. It accords with the statistic perspective that training dataset from more days contains more details about every hand motion pattern.

5. It is also suggested that the relatively effective choice of number of days for training is three, because the recognition results improve obviously when the number added into three for most of subjects in our experiments and the recognition performance change slightly when the number is specified as 3, 4, or more.