Neural network-based shape recognition using generalized differential evolution training algorithm

In this paper a new method for recognition of 2D occluded shapes based on neural network using generalized differential evolution training algorithm is proposed. Firstly, a generalized differential evolution (GDE) algorithm is introduced. And this GDE algorithm is applied to train multilayer perceptron neural networks. Then a new shape feature, refer to as multiscale Fourier descriptors (MFDs) is proposed. Finally, the superiority of GDE training method over traditional approaches to train networks is demonstrated by experiment. The experimental results show that our proposed GDE training method is much efficient and effective. And they also showed that the MFDs method is suitable for the shape recognition.