基于改进的PSO-SVM的音频信号特征识别和分类

针对音频信号准确性分类的问题,提出一种基于改进的的粒子群优化算法(PSO)的支持向量机(SVM)音频信号分类的方法,简称IPSO-SVM.首先用Mel倒谱系数法对4种音频信号进行特征提取.其次在PSO中引入自适应变异因子,能够成功地跳出局部极小值点;然后对PSO中的惯性权重进行了改进,将惯性权重由常数变为指数型递减函数.随着迭代的进行,使权重逐渐减小,这样做有利于粒子进行局部寻优.最后用改进的PSO不断优化SVM中的惩罚因子c和核函数参数g来提高预测精度.实验结果表明,与传统的SVM、PSO-SVM、GA-SVM相比,我们提出的IPSO-SVM算法分类结果更精确.

Feature Recognition and Classification of Audio Signals Based on Mproved PSO-SVM

In this paper,aiming at the problem of classification of audio signals,a new algorithm based on Improved Particle Swarm Optimization (PSO) is proposed for the classification of audio signals of the support vector machine (SVM),which is called IPSO-SVM.Firstly,4 kinds of audio signals are extracted by using Mel frequency spectrum coefficient method.Secondly,the adaptive mutation factor is introduced into PSO,which can successfully jump out of the local minimum point.Then,the inertia weight of PSO is improved,and the inertia weight is changed from the constant to an exponential decreasing function.With the iteration,the weight is gradually reduced,so that it is advantageous to the local optimization of the particle.At last,the improved PSO is used to optimize the penalty factor C and kernel function parameter g in SVM to improve the prediction accuracy.Experimental results show that compared with the traditional SVM,PSO-SVM and GA-SVM,the classification results of IPSO-SVM algorithm are more accurate.