改进的支持向量机SMO算法说话人识别系统研究

Speaker Recognition Research Based on Support Vector Machine of improved SMO algorithm

支持向量机是统计学习理论的一个重要的学习方法,也是解决模式识别问题的有效工具。本文把支持向量机应用在说话人识别系统中,对支持向量机的SMO算法进行了论述,并对SMO中有关两个待优化拉格朗日乘子的选取做了改进,用简单的排列算法取代函数集中的遍历操作来使目标函数值下降,实验证明SMO算法具有占用内存少,运算速度快等优点,本文中的SMO改进算法可以节省50%时间。

Support Vector Machine （SVM） is a new and very promising classification technique. The approach is properly motivated by statistical learning theory. It is a strong tool to solve identification problems. This paper studies the speaker recognition applied in Support Vector Machine, discuss the SMO algorithm, and improve the selection of the two preparation optimization lagrange multiplier in SMO relevant to pick two laplacian multiplier for optimization, using the sequencing array instead of alternating loop to decrease the target function value, the experiment certificates that SMO algorithm really has the advantage of taking up little memory and fast calculate speed, The improvement to SMO algorithm in this paper is proved that it can save time 50% approximately compared with pre-improved situation.