Gain parameters based complex-valued backpropagation algorithm for learning and recognizing hand gestures

In this paper, an improved complex-valued backpropagation algorithm with gain parameters is proposed. It is then employed to train a complex-valued feedforward neural network with one hidden layer. The well-trained complex-valued neural network is finally applied to deal with the recognition problem of 26 hand gestures. The results of experiment clearly show that much better performance can be achieved by our improved complex-valued backpropagation algorithm than some existing methods.