Research On LMS Channel Estimation Algorithm For MIMO-OFDM Systems

With the rapid development of the mobile telecommunication and the WLAN, it is necessary to use anti-fading techniques on limited frequency resources to achieve high-speed data transmission and multimedia services. Orthogonal Frequency Division Multiplexing (OFDM) is a special multi carrier transmission technique with anti-multipath transmission and anti-frequency-fading characters. Multi-In-Multi-Out (MIMO) is a significant breakthrough in wireless mobile telecommunication area which can improve the capacity and frequency utilization. Combining OFDM and MIMO together as the MIMO-OFDM systems, has been a keynote in the next generation mobile telecommunication techniques.The paper based on the analysis and summarization of the relative documents, introduces the history and situation of the OFDM technique. With the analysis of several classic wireless channel transmission models and the study of the MIMO-OFDM system’s theory, the paper uses Matlab simulation to test and verify the MIMO system capacity’s changes when the transmit and receive antenna’s amount changes. The paper also studies the OFDM channel estimation techniques by finishing the simulations and comparisons of the LS, LMMSE and downward rank LMMSE algorithms as well as several classic interpolation algorithms such as linear, Gaussian, Cubic-spline and Lagrange algorithms.Although LMMSE and MMSE algorithms are with good simulation precision, but their high complexity make them unpractical for applications. LS algorithm is of low complexity; however, its low precision is critical for channel estimation, making it not a ideal method for channel estimation. Although it is probable using the combination of LMS and LS to achieve accurate channel estimation, but LMS’s rate of convergence is slow.Therefore, this paper puts forward a modified LMS algorithm which can adaptively changes the step length to upgrade the speed of the LMS’s rate of convergence. Simulation results shows that the LMS algorithm has a higher simulation precision than the LS and a higher rate of convergence than the LMS.