一种基于Fisher准则的有监督表情识别算法

Supervised expression recognition algorithm based on Fisher criterion

在分析LPP算法存在的不足后,提出了一种新的基于Fisher准则的有监督保局投影表情识别算法,即FSLPP。该算法通过可调因子有效地结合人脸局部流形的结构信息和样本的类别信息,对表情图像序列提取其Gabor特征后采用FSLPP算法获取低维表情特征序列,并由SVM分类器估算识别率。在JAFFE人脸表情库对该算法进行了测试,结果表明,与FLD、LPP等方法相比,该方法具有较好的识别率。

After analyzing the shortcomings of LPP algorithm,this paper proposed a new expression image feature extraction and recognition method based on supervised locality preserving projections and Fisher criterion（ FSLPP） . The algorithm effectively integrated the face manifold local structure information with the labels’information by adjustable factor. Extracting human facial expression characteristics by Gabor wavelet,extracted the low-dimensional feature of expression for recognition by FSLPP algorithm,used the support vector machine （ SVM） algorithm to construct classifiers. The proposed method was tested and evaluated using the JAFFE face expression database. Experimental results show that FSLPP is more powerful than Fisherface and Laplacianface for expression feature extraction and recognition.