Abstract: In the long term evolution (LTE) system, accurate channel estimation (CE) is crucial in channel equalization and coherent demodulation. Due to the discrete non-uniform pilot pattern of the LTE downlink system, the traditional CE method results in significant estimation errors over high-speed mobile channels. In order to improve the CE accuracy, firstly, the discrete Fourier transform (DFT)-based CE technique is developed to enhance the CE accuracy of pilot. Then according to the pilot pattern of the LTE system, we propose two CE methods to improve the CE accuracy of data: 1) When the statistical properties of channel are unknown, the non-uniform pilot pattern is transformed into the uniform pilot pattern. Then the spline fitting is utilized to estimate the channel frequency response (CFR) of data position. 2) When the statistical properties of channel are known, a high performance adaptation interpolator is designed based on the minimum mean square error criterion, which can accurately obtain the required coefficients of CE. Simulation results and analysis demonstrate that, compared to the conventional linear interpolation algorithm, the proposed schemes can exploit more pilot information, which significantly improve the CFR accuracy and enhance the demodulation performance.