Font Size: a A A

Study On The Key Technologies Of Compressed Sengsing Based On Structural Characteristics Of Signal

Posted on:2016-11-17Degree:DoctorType:Dissertation
Country:ChinaCandidate:W W LiFull Text:PDF
GTID:1108330482457710Subject:Communication and Information System
Abstract/Summary:PDF Full Text Request
Compressed Sensing or Compressive Sampling (CS) theory brings new revolutionary breakthrough for signal compression coding technology. Compared with the traditional compression sampling algorithm, CS algorithm has great potential to reduce the data acquisition and processing cost, save storage space and improve the transmission efficiency. The selected topic has important theoretical significance and broad application prospects.Designing the effective CS algorithm to improve the quality of signal reconstruction is an essential condition of successfully promoting and applying the theory for actual data compression sampling system. It is the main goal of this thesis. The research work of this thesis is mainly to design good CS algorithm for further enriching the CS theory by exploiting the structural characteristics of the signal. The main research content and innovation points of this thesis include:1. For the problem that sparse representation based on wavelet transform leads to part of the image information has been lost before the measurement sampling, a block compressed sensing (block-CS) scheme based on the image block similarity is proposed. In order to improve the signal sparse degree and reduce the loss of information, classification algorithm is proposed by utilizing the image block similarity. The simulation results demonstrate that the proposed algorithm can effectively improve the quality of the reconstructed images.2. In order to further improve the performance of the block-CS algorithm, a clustering compressed sensing (CCS) algorithm is proposed based on the image block similarity. The classification algorithm is introduced into the CS framework by utilizing the image block similarity. And the most optimal image block in each class as representatives to be transmitted, rather than transmitting all the image blocks. The proposed method decreases the amount of data transmission. In addition, for further improving the performance of the CCS algorithm, an unequal-CCS algorithm is proposed based on the characteristics of wavelet coefficients. Simulation results show that, compared with existing block-CS algorithms, the proposed algorithm can significantly improve the quality of reconstructed images.3. For making the general image signal can also be used in multiple measurement vectors (MMV) model, this thesis designs a structured CS algorithm based on characteristics of image wavelet coefficient. Compared with CS algorithm based on single measurement vector (SMV) model, the recovery performance of CS algorithm based on MMV model is better. Firstly, zero insertion algorithm is proposed by utilizing quadtree representation of wavelet coefficients, and then the SMV model can be convert into the MMV model such that the MMV model can be used for general images rather than only several special signals. In order to further enhance the precision of image reconstruction, the classification algorithm is introduced into the MMV framework by using the image blocks similarity. In this way, the number of the columns of joint sparse matrix is increased dramatically. The proposed algorithm can significantly improve the accuracy of image reconstruction under the same number of measurements.4. In order to solve the application limitations of one dimensional signal in MMV model, this thesis proposes a structured CS algorithm based on one dimensional signal waveform similarity. Firstly, the segmentation algorithm is proposed by using the signal waveform similarity, and then the one dimensional signal is converted into a two-dimensional joint sparse matrix. It makes the signals can be used in MMV model. Numerical experiment results show that the proposed algorithm can achieve high quality recovery. More importantly, compared with the CS algorithm based on the SMV model, the performance of MMV model transformation algorithm is more stable under the condition of the larger compression ratio (CR).5. This thesis makes a preliminary study on the CS theory in the application of joint source channel coding (JSCC). An unequal error protection (UEP) scheme is proposed based on the CS theory. The proposed algorithm not only applies the CS algorithm on source coding to realize the effectiveness of the wireless communication system, the CS method is applied on channel coding to ensure the reliability of wireless communication system simultaneously. The experimental results show that the proposed scheme can achieve high peak signal-to-noise ratio (PSNR), and significantly improve the fault-tolerant performance of the system. In addition, the proposed CS-UEP scheme has good robustness, especially suitable for worse channel condition.Finally, the whole research work is summarized, and research on the key technology of CS theory based on the structural characteristics of the signal is prospected.
Keywords/Search Tags:Compressed Sensing, Wavelet Transform, Similarity, Single Measurement Vector, Multiple Measurement Vectors
PDF Full Text Request
Related items