Study Of Frequency-domain Block LMS Adaptive Filtering Algorithm

The LMS (Least Mean Square) adaptive filtering algorithm was firstly proposed by Widrow et al. It is widely used in adaptive control, radar, system identification and other areas for its advantages, which are simple structure, low computational complexity and high stability. However, there are some flaws in traditional time-domain LMS adaptive filtering algorithm, including two aspects:the one is the conflict on requirement for convergence factor between its convergence speed and steady-state error, so a compromises should be made between the high convergence rate and the low static error when choosing the convergence factors; the other is the convergence performance is affected by the input signal power. So, under the changing of the input signal power, in order to ensure the convergence stability of the maximum input power, the convergence rate of low input signal power must be sacrificed. Besides all the deficiencies above, for any signal processing system, particularly the system requiring real-time processing, reducing the computational complexity of algorithms has been an important direction for improvement.In this paper, we start with the research for the significance of adaptive filtering technology. Firstly the algorithm performance of traditional time-domain LMS, RLS and the block LMS have been studied. Secondly a block-based frequency-domain LMS adaptive filtering algorithm was introduced and analyzed based on the traditional time-domain algorithm. Then the system model of the traditional time-domain LMS and frequency-domain block LMS adaptive filtering algorithm was built. Finally using Verilog hardware description language programming, the simulation examination for traditional time-domain LMS and frequency-domain block LMS adaptive filtering algorithm were completed in the platform of Xilinx ISE Design Suite 10.x, while two algorithms were analyzed in the size of computational and the ease of hardware implementation.