Skip to Main Content
Facial affection recognition performance can be improved by the combination of Gabor wavelet transformation and fractal dimension and the classification by BP neural net. Preprocessing to grey images as grid division are firstly completed. Feature parameters which indicate the property of facial affection are achieved through the transformation by Gabor wavelet and the computation of box-counting dimension and differential fractal dimension. When applied in recognizing four, six and seven types of facial affection of the same person, the algorithm can reach the recognition performance of 87%, 78%and 77% respectively. The performance of the algorithm is improved with the increasing of the type of affection.