基于一类SVM概率密度估计的多分类贝叶斯算法研究

为降低训练分类器的运算复杂度，并解决支持向量机（SVM）对多类分类问题没有特别有效解决方法的问 题。提出了一种基于一类支持向量机的多分类贝叶斯算法，证明了基于径向基核函数的一类SVM的分类函数归 一化为密度函数，并将所得的概率密度函数用于构造二分类及多分类贝叶斯分类器。仿真实验将提出的多分类贝 叶斯算法应用于多类通信信号调制识别，结果表明：该算法的分类准确率不低于传统SVM多分类器，而在多类属、 每类训练样本数目较大的情况下训练所需的运算量和存储量仅是传统SVM多分类算法的0.5%大大减小了核 矩阵规模和

Research of multi-class Bayesian algorithm based on one-class SVM probability density estimation

For reducing the computational complexity in training classifier and solving the problem of support vector machine (SVM) to multi class classification, a one-class SVM based multi-class Bayesian classifier is proposed. It is proven that the solution of one class SVM using the Gaussian kernel can be normalized as an estimation of probability density, and the probability density can be used to construct the two-class and multi-class Bayesian classifier. The proposed classifier is used to recognize the modulation scheme of multi communication signals. Experimental result showed that the correct classification probability of the proposed classifier is comparable to traditional multi-class SVM classifier. In the condition of large class amount and large amount of training samples of each class, the calculation amount of training and storage is only 0.5 percent of the traditional SVM classifier, thus the size of kernel matrix of the new algorithm is greatly less than traditional multi-class SVM, which lead to less training time for the new classifier.