Research On Channel Estimation Algorithm Based On Improved Threshold In OFDM Systems

Orthogonal Frequency Division Multiplexing is a technology of multi-carrier modulation and multiplexing. With the advantage of robustness against multi-path delay spread and high band width efficiency, OFDM technology is widely used in field of high-speed data transmission. Channel estimation has a direct impact on the performance of OFDM system, moreover, channel estimation is one of the research hotspot in OFDM system, but also the fundament of coherent detection, demodulation and equalization.Firstly, the thesis introduces the radio channel transmission characteristics, focusing on the principle of multipath propagation and Doppler effect. The thesis elaborate on the principle of OFDM system, include the principle of OFDM modulation and demodulation, the effect of the guard interval and cyclic prefix, and how to select parameters of the OFDM system. On this basis, focusing on the typical pilot-based channel estimation algorithms, and analyze the performance and complexity of the various algorithms.Secondly, the DFT channel estimates based on threshold was carefully studied in this thesis, A new DFT channel estimation algorithm, which takes the median of the energy of all the time domain noise samples point as threshold, was proposed by analyzing shortcomings and deficiencies of these existing thresholds. Simulation results showed that the proposed algorithms can effectively solve the problem that the useful channel samples are filtered under big impulse noise, suppress noise inside the CP effectively, make the bit error rate and mean square error further decrease, and improve the accuracy of channel estimation.Finally, the channel estimation using wavelet de-noising was studied in depth. The study found the channel frequency response obtained by least squares channel estimation has a certain discontinuity or rapidly changing point, so when using wavelet de-noising,there will be a pseudo-Gibb phenomenon, therefore, the translation invariant wavelet de-nosing was applied to channel estimation and propose a improved threshold function to handle the high-frequency coefficients of wavelet. Simulation results showed that the proposed algorithms can effectively suppress the pseudo-Gibbs phenomenon and filter out the noise, make the bit error rate and mean square error further decrease, and get a more accurate channel frequency response finally.