| The Long Term Evolution (LTE) is a new research and development project launched by 3rd Generation Partnership Project (3GPP) organization for the next generation communication systems. The LTE systems takes orthogonal frequency division multiplexing and single-carrier frequency domain equalization as core technology. Channel estimation is of great significance in broadband communication systems, and it is necessary to study the channel estimation algorithm specially for LTE systems.This paper studies channel estimation method applicable for LTE uplink. We introduce two basic criterion, namely Least Square (LS) and Linear Minimum Mean Square Error (LMMSE), and the simulations show that LMMSE can achieve better performance than LS, but is more complex. The estimation accuracy can be greatly improved by truncating the LS results in time domain, according to the length of channel.It's block pilots that will be used in LTE systems, and block pilots has inherent deficiencies of resisting Doppler frequency shift. In this paper, we use polynomial fit and interpolation method to enhance the performance of systems under high Doppler spread. By simulation, we give the MSE results for different fit parameters, under twelve kinds of channel environment, which is made up with three sets of channel parameters and four SNRs. Under the principle of lower complexity and higher performance, we choose the best fit parameters for each channel environment, forming an adaptive polynomial fit and interpolation strategy. The simulation shows that high order fit applies only for high Doppler spread and high SNR. This strategy can make a better balance between performance and complexity.The estimation accuracy can be further improved by exploring the time domain correlation of the channel, that is, passing the initial estimation results in frequency domain through a time domain filter. We analyze the Doppler-resisting performance of time domain Wiener filter. According to the principle of normalized least mean square (NLMS) filter and recursive least square (RLS) filter, the method of adaptive channel estimation is derived. By simulation, we analyze the impact of different filter parameters on NLMS and RLS algorithm, and the convergence properties of the two. Simulation shows that RLS achieves better Doppler-resisting performance than NLMS, and the Doppler spread and the length of filter has no obvious influence on the speed of convergence. |
