Research On Compressive Sensing In Wireless Communication Systems

As a new low speed sampling technology, compressive sensing break through the limitation of the traditional Nyquist sampling theorem and has attracted much attention in the field of communications. This thesis mainly disscuses two applications of compressive sensing for large multiple input multiple output (MIMO) beam vectors searching and ultra-wideband channel estimation.In multiple antenna system, beamforming technology is usually applied to reduce path loss, improve signal to noise ratio and improve the system performance. In beamforming, the key issue is to find the optimal beam vectors. Non-codebook beamforming usually needs a series of process to find the optimal beam vectors: channel estimation, channel decomposition and feedback from receiver to transmitter. When the number of elements in array antenna is large, the training cost is huge.Based on the characteristics of large antenna array MIMO channel, we observe that the steering vectors constituted by different angles are approximately orthogonal, thus, channel matrix can be expressed in the form of approximate singular value decomposition (SVD). By using steering vectors instead of singular vectors, the tough task of channel singular value decomposition is transformed to identify the steering vectors corresponding to strongest path in given dictionary. In time division duplexing system, taking advantage of sparsity of angles, we propose a matching pursuit based singular vectors estimation scheme. With low complexity, the proposed scheme acquires optimal beam vector pairs without tedious channel estimation or the SVD decomposition. By using iteration strategy, this scheme can achieve better performance compared with the traditional direct estimation algorithm and power iterative algorithm.In the ultra-wideband wireless communication system, in order to ensure the quality and reliability of communication, the receiver must handle the channel information. Therefore, the channel estimation is a key technology of ultra-wideband wireless communication system. However, due to wide bandwidth, the sampling rate is even more than GHz, which is formidable for analog to digital converter and digital signal processor.This thesis analyzes the statistical sparsity within real ultra-wideband signal. We construct a new eigen dictionary by eigen vectors, in this way, the problem of channel estimation can be converted to the problem of compressive sensing. Then, taking advantage of the correlation between the hyperparameters, we propose two methods based enhanced Bayesian compressive sensing for ultra-wideband channel estimation. Scheme 1 achieves the channel estimation by means of cyclic updating among hyperparameters, expansion coefficient mean and covariance. To avoid matrix inversion process, in scheme 2, intermediate variables are suggested. Besides, scheme 2 uses the changes of logarithm likelihood function as terminate criterion for iteration. Both schemes can efficiently recover ultra-wideband signals from far fewer measurements. Moreover, multi-task Bayesian compressive sensing developed by single-task version can obtain further reconstruction performance.