摘要
针对混合核支持向量机(SVM)中的可调参数一般是根据经验或人工随机调试得到,不能确保参数最优的局限性,提出用粒子群和人工蜂群的并行混合优化(ABC-PSO)算法来优化混合核SVM参数,找出满足条件的最优参数组合.将该SVM模型应用到语音识别中,通过对三个不同语种的语音数据库的实验仿真,验证了混合算法优化SVM参数所得的优化SVM模型比PSO算法优化SVM所得的模型,具有良好的泛化能力和语音识别能力.
The parameters of support vector machine （SVM） with mixed kernels were used to be gotten manually by experience which has limitations that it could not be the best one. A method is put forward by using a parallel hybrid algorithm of particle swarm and artificial colony bee （ABC-PSO） to optimize the parameters of mixed kernels SVM. The method is used to find out the optimal parameters combination which meets the conditions in speech recognition. Through simulation experiments of three databases show that the SVM with obtained parameters optimized by the hybrid algorithm has better generalization and speech recognition ability than the SVM optimized by PSO.
出处
《数学的实践与认识》
CSCD
北大核心
2014年第18期158-165,共8页
Mathematics in Practice and Theory
基金
国家自然科学基金(61072087)
太原科技大学博士基金(20142003)
