【摘要】 传统图像特征提取具有较高维度缺陷,造成算法分类效率低、复杂度高、分类速度慢、计算开销大等问题。为此提出AAM算法,定位关键点提取人脸表情几何特征。将朴素贝叶斯分类器结合特征属性重要度调节高斯核函数,使用K近邻算法实现分类决策,提出一种WNBC-KNN分类方法,从降低数据维度和分类算法两方面优化人脸表情分类。在CK+数据和JAFFE数据集上实验,识别率分别达到90%和86%。与传统的朴素贝叶斯算法比较,改进后的算法识别率分别提高6%和30%。更多还原

【Abstract】 The traditional image feature extraction algorithms have many disadvantages,such as low classification efficiency caused by the high dimensionality,high algorithm complexity,slow classification speed and high computational overhead. This paper proposes an algorithm that extracts the geometric features of facial expression based on the AAM to locate key points. Combining Na?ve Bayes classifier with feature attribute importance to adjust the Gaussian Kernel function,and using KNN(K-Nearest-Neighbors)algorithm to implement classification,this paper introduces the WNBC-KNN classification algorithm. The WNBC-KNN classification algorithm optimizes facial expression classification from two aspects:data dimension reduction and the classification algorithm. This paper achieves90% and 86% recognition rate on CK + data and JAFFE data set. Compared with the traditional Naive Bayes algorithm,the proposed algorithm has improved the recognition rate of CK + dataset and JAFFE dataset by 6% and 30%.更多还原