Research On1-bit Compressive Sensing Technology

Compressive sensing(CS) utilizing the signal characterisitic of sparsityovercomes the limition of Nyquist sampling theorem. The compression andsampling can be performed simultaneously. In recent years, CS has developedrapidly, It is widely available in many fields, such as analog information conversion,optical imaging, military radar and aerospace. With the developing of CS, therehave been many emerging branches.1-Bit compressive sensing, an importantbranch of CS, has attracted widely attention for its simple framework and perfectreconstruction performance.1-Bit CS has a wide promising application, includingthe fields of wireless communication and cogniitive radio.1-Bit CS can accurately recovery the signals even when each measurement isquantized to just1-Bit. This property enables the design of simplified hardwarestructure, reduction of the space taken by samples and enhancement the speed ofdata saving and transmitting. While this direction has been late to get started, mostof the work was limited in theory. Most algorithms of1-Bit CS denpen on signalsparsity and the research of1-Bit CS implementing technology is lack. This thesisfocuses on the problems of1-Bit CS. The main research contents and results of thethesis are listed as follows:1、The basics of1-Bit CS are researched. The1-Bit quantization model, theconstraint condition of measurement matrix and the existence and sparsity of thesolution are decribed first. And then we discuss the Binary Iterative HardThresholding (BIHT) algorithm, which is an outstanding1-Bit CS recoveryalgorithm. We introduce a series of extension algorithms of BIHT, and provide theirmodel and the implementation steps. The simulations validate that the AdaptiveOulter Pursuit (AOP) algorithm and AOP-flips algorithm have better performancethan other algorithms, when measurements contaminated with sparse nosie.2、In face of the difficulty of getting the sparsity level of signals in practical,a blind1-Bit CS algorithm(Sparse Adaptive Binary Iterative Hard Thresholding,SABIHT) without prior information of the sparsity is proposed based on the BIHTalgorithm. After analysis of the sparse dependence problem of BIHT algorithm, anew algorithm, which adopts sparse adaptive strategy, is introduced. In thenumerical experiments, the results show that the SABIHT algorithm can achieve performance as good as the BIHT algorithm without prior information of thesparsity.3、In face of the absence of hardware prototype based on1-Bit quantizating,we study the sampling systems based on1-Bit quantization. First of all, theframework of multiple tone sparse signal is researched. Then, we put forward adesign scheme of sample system basic on1-Bit quantization for sparse multipletone signals, which includes a multiplier, a filter, a sampling holder and acomparator. The structural scheme of sensing matrix is described. We complete thesystem design and carry out experments on the system. Experimental resultsdemonstrate that the proposed scheme for multi-tone sparse signals can realizestable and successful reconstruction. In the condition of fixed bit-budget,reconstruction with1-Bit quantization samples obtains better recoveryperfermoance than those with more-Bit quantization.4、The sample system based on1-Bit quantization for multi-band signals isproposed. We describe the multi-band signal model and the architecture of theproposed system. The signal is multiplied by a periodic waveform and filtered by alowpass filter. The filtered is compared with a fixed value to get1-Bit samples atuniform time instants. A corresponding algorithm named simultaneous binaryiterative hard thresholding-2is designed for the proposed system. In simulationswe compare our system with modulated wideband converter to prove its advantagesin the condition of constrained Bit-budget, particularly in low input signal to noiseratio levels. The proposed system shows some performance gain in engineeringaspect: robustness to noise and mismodeling, potential hardware simplifcations,real time performance for signals with time-varying support and stability toquantization effects.