一种基于SVM的声源定位算法

A Sound Source Localization Algorithm with Support Vector Machine

随着多媒体技术的快速发展,获取高质量的语音成为一种越来越受到广泛重视的技术手段,以麦克风阵列定位声源的方法在诸多领域得到了广泛的应用。支持向量机(Support Vector Machine,SVM)是一种基于统计学习理论和结构风险最小化原则的机器学习方法,其诸多参数的选择直接影响到SVM的性能,SVM可作为估计声源位置的分类器,且可通过选取适当的参数提高算法的抗噪声能力。为此,提出了一种基于SVM的声源定位新算法。该算法提取鉴别互相关函数的特征,通过选取合适的参数,对SVM的核函数进行优化。基于Matlab对提出的新算法进行了仿真实验验证。仿真实验结果表明,该算法较为显著地增加了混响和噪声条件下声源定位的准确性,且具有良好的鲁棒性。

With the rapid development of multimedia technology, the method to obtain high-quality voice has become an increasingly widespread technical means, which has been widely used for microphone array sound source localization in many fields. Support Vector Machine （SVM） is a machine learning method based on statistical learning theory and structural risk minimization principle, and its performance depends on the correct selection of related parameters. SVM can be used as a classifier to estimate the position of the sound source, and it can improve the anti-noise ability of the algorithm by selecting the appropriate parameters. Therefore a new kind of sound source localization based on SVM is proposed. By extracting the features of cross-correlation function and selecting appropriate parame- ters, the kernel function of SVM is optimized. The simulation has been carded out with the proposed algorithm using Matlab, results of which show that it has increased the accuracy of sound source localization significantly in noisy and reverberant environments, with good robustness.