摘要
基于多通道到达时间差(TDOA)的定位方法是声源到达方向(DOA)估计中的重要方法。其中,由TDOA到DOA的映射是该方法的一个关键,目前广泛采用的映射方法为最小二乘法。然而最小二乘法存在诸如声源位于阵列端射方向时性能急剧下降的问题。为克服这一缺点,提出了基于最小二乘支持向量回归机(LS-SVR)的映射方法。在支持向量机技术中,核函数的选取直接影响着支持向量机的性能,但之前的工作仅讨论了径向基核函数。针对声源DOA估计中的TDOA映射问题,研究了径向基核、多项式核以及线性核函数构造的LS-SVR对声源DOA估计的影响。
The technique based on time -differences- of- arrival （TDOAs） with multiple channels is an important method for the sound sources direction -of- arrival （DOA） estimation. One of its key technologies is the mapping of TDOAs to DOA. At present ,the least - squares based algorithm is widely used in sound sources DOA estimation. It has been noted that the perform- ance of the least - squares algorithm is deteriorated greatly around the end - fire directions. To combat this problem, a mapping method based on least- squared- support- vector- regression （IS- SVR） is proposed ,where is used radial basis kernel func- tion. The choice of kernel functions plays an important role in the SVM applications. The mapping construction of IS - SVR with radial basis kernel ,polynomial kernel and linear kernel function is focused,which influence the sound sources DOA estimation.
出处
《电声技术》
2014年第5期57-61,共5页
Audio Engineering
基金
国家自然科学基金项目(61001150)
江苏省自然科学基金项目(BK2010495)
