Compressed Sensing Channel Estimation Based On Golay Sequences

Orthogonal Frequency Division Multiplex(OFDM) has many advantages in communication just as its high spectrum efficiency and effective resistance to channel delay and multipath fading, so that it is widely utilized in 3G and 4G system. In the OFDM system channel estimation is necessary for the purpose of signal coherent detection, and the estimation precision will directly influence the performance of the communication system, therefore channel estimation is listed as one of the three key technologies of OFDM system. Currently, existing channel estimation methods assume that channel is intensive and of many paths, but actually channel has sparse property in real communication system, which is to say only small parts of channel tap coefficient are non-zero.In 2006, a novel signal processing method known as Compressed Sensing has been proposed by Donoho[1-2] and Candes[3]. In this theorem, if the signal is sparse or can be compressed, the sampling rate can be fairly lower than Nyquist rate, then based on optimization theorem, the original signal can be reconstructed with a high probability. Currently in compressed channel estimation, random sequence is normally uesed as training sequence. But it will cause a large peak-to-average ratio in the OFDM system, also its difficulties in hardware implement and high computational burden in original signal reconstruction.Based on the problems from previous literature, in this thesis, Firstly, we analyzes whether the wireless channel is sparse or not, in the purpose of compressed sensing take into the channel estimation is reliable. According to the principle of communication by the knowledge of convolution, we can set up a channel estimation model, afer all we analyze how to using the training sequence for constructing a measurement matrix with toeplitz structure. For its perfect auto-correlation, timing synchronization and low peak-to-average power ratio, based on compressed sensing theory, deterministic Golay sequence is utilized to estimate sparse channel as training sequence. At the same time, the performance impact in estimated value for channel impulse response compared with that of random Gaussian sequence, and compare the computational complexity of the two sequences in reconstruction of channel impulse response. Finally, using Golay sequences as the training sequences, At the receiver oversampling the signal establish multipath time delay estimation in chips based on compressed sensing.Simulation result indicates that both Golay sequence and random Gaussian sequence can reconstruct non-zero tap coefficient of sparse channel. But mean squared error of Golay deterministic sequence is less than random Gaussian sequence in channel impulse response estimation of sparse channel. Also, Golay sequence can limit Peak-to-Average ratio with 3dB with computational complexity less than random Gaussian sequence in channel impulse response reconstruction. The multipath time delay estimation based on compressed sensing, under certain signal to noise ratio and the sampling points can accurately estimate the multipath time delay.