滑动DCT-LMS自适应算法研究与FPGA实现  

Research and realization of sliding DCT-LMS adaptive algorithm based on FPGA

作　　者：高海飞[1] 张浩[1] 侯富诚[2] 向永波[1] 

机构地区：[1]中国科学院微电子研究所,北京100029 [2]中国科学院苏州纳米技术与纳米仿生研究所,江苏苏州215123

出　　处：《传感器与微系统》2012年第5期13-15,19,共4页Transducer and Microsystem Technologies

基　　金：国家"863"计划资助项目(2009AA011606)

摘　　要：研究一种改进的低复杂度复数滑动离散余弦变换(DCT)最小均方(LMS)自适应算法,并设计该算法的FPGA实现结构。在常规LMS算法的输入端前添加改进的滑动DCT,降低输入信号之间的关联性,提高自适应算法收敛速度。改进的滑动DCT算法针对硬件实现进行了优化,提高其在硬件实现中稳定性和精度。给出算法在FPGA实现框图、结果和Matlab仿真结果的对比,以及算法在FPGA中的资源使用。算法已经在实际工程中应用,效果远优于常规LMS自适应算法。A low-complexity complex-valued sliding discrete cosine transform least-mean square（DCT-LMS） adaptive algorithm and its realization structure on field programmable gate array（FPGA） are researched.A revised sliding DCT is added in input port of conventional LMS algorithm to guarantee a fast convergence rate when relevance of input signals are reduced.The revised sliding DCT algorithm is optimized to improve stability and precision in hardware realization.The realization structure,result and the comparison with Matlab simulation result are presented.The resource utilization of the algorithm in FPGA is analyzed.The algorithm has been successfully used in practical engineering and its effect is better than conventional LMS adaptive algorithm.

关 键 词：离散余弦变换 最小均方 自适应滤波 FPGA实现 

分 类 号：TN911.7[电子电信—通信与信息系统]