摘要
针对多重信号分类(multiple signal classification,MUSIC)算法计算量大不适于实时处理的问题,提出了一种并行处理方案。首先,根据协方差矩阵的Hermite特性简化其构造过程;再通过实值化预处理,将后续运算转换到实数域,通过Householder变换将协方差矩阵简化为三对角矩阵,对三对角矩阵进行QR分解得到特征值和特征向量用于谱峰搜索;最后,各个阶段都适于采用多处理器并行处理。通过理论分析和仿真,验证了该方法在对MUSIC算法性能影响不大的前提下能大大减小运算量,提高算法处理速度,为MUSIC算法的高效化实现提供了一定的理论基础。
The computing load of the multiple signal classification（MUSIC） algorithm is large,thus it is not suitable for real-time processing.A parallel processing scheme is proposed to solve this problem.The construction of a covariance matrix can be simplified according to its Hermite characteristics;and by the real-value preprocessing,the sequential operations are converted to the field of real numbers.Then the covariance matrix is simplified as a tridiagonal matrix by using Householder transformation,and the eigenvalue and eigenvector of the tridiagonal matrix obtained by QR decomposition are used in spectral peak searching.Finally,the multiprocessor parallel processing technology is fit for each stage of the algorithm.Theoretical analysis and simulation results prove that this method reduces the computing load greatly and increases the speed of processing with little impact on the performance of the algorithm,and it provides a theoretical basis to the efficient realization of the MUSIC algorithm.
出处
《系统工程与电子技术》
EI
CSCD
北大核心
2012年第1期12-16,共5页
Systems Engineering and Electronics
基金
中国博士后科学基金(20070410262)资助课题
关键词
阵列信号处理
并行处理
空间谱估计
array signal processing
parallel processing
spatial spectrum estimation
