Abstract:

In order to find a gesture recognition method with fast and high recognition rate, a gesture recognition algorithm is presented based on an improved probabilistic neural network. The improved algorithm uses K-W method to filter the most representative features of sEMG（SurfaceMyoelectrogram Gestures） features, and utilizes particle swarm optimization method to optimize the transmission rate. In the experiment of identifying seven kinds of hand gestures, the average correct recognition rates of improved probabilistic neural network are more than 90%, while the traditional BP algorithm was only 85.% correct. Simulation results show that the improved neural network algorithm has much shorter training time and stronger classification ability.

Key words: surface-myoelectrogram gcestuses（sEMG）, pattern recognition, probabilistic neural networks, K-W test, particle swarm optimization