抄録
Recently, recognizing a human expression plays an important role in a machine cooperating with people. This paper proposes a method to recognize expression in facial images using novel features. First, facial key parts are extracted by using a generic Active Appearance Model (AAM), then Gabor magnitude pictures are obtained by convolving the images of facial key parts with Gabor filters, followed by encoding with Local Directional Pattern (LDP) operator which enhances facial local feature. The maps generated by these processes are divided into several blocks, and Local Gabor Directional Pattern Histogram Sequence (LGDPHS) based on facial key parts are extracted from each block. The histograms are concatenated to a vector, where Principal Component Analysis (PCA) is used to reduce the dimensions. Finally, the feature vector is classified by Support Vector Machine (SVM). In our experiment, we demonstrate both person-independent and person-dependent facial expression recognitions (FERs), and the recognition rate of our person-independent FER improves up to approximately 16% as compared to the conventional methods. The performance of our recognition reaches to 94.74% in person-dependent FER. As a result, we have successfully demonstrated effectiveness of our proposed method.
