A feature fusion base perceiving emotion method and a system thereof for recognizing the feeling of the human through the speech signal and face video signal are provided to perform emotion recognition by extracting information from voice signal and image signal. A first information is extracted from collected voice signal corresponding to acoustic feature(P1, P2). A second information including a mouth, eyes, and eyebrows are extracted from facial image signal in which is second information is recognized(P3, P4). A feature value is selected in the first information and the second information by using SFS(Sequential Forward Selection) method(P5). A pattern by emotion is classified by seeing up multi-layer perceptron as an input(P6).
展开▼
