Research On Bayesian Reconstruction Methods For Quantized Compressed Sensing Of Odor Signals

In the process of rapid socio-economic development and the construction of a harmonious society,environmental security has become a bottleneck restricting the rapid economic development and building a harmonious society.Leakage of toxic odors is one of the most serious accidents in environmental pollution accidents.The necessity and effectiveness of using sensor network technology to monitor the leak status of toxic odors and locate the source of leakage are gradually recognized by the engineering.In the wireless sensor network for toxic odor leakage monitoring,the odor sensor following the Nyquist sampling theorem has high sampling rate,a large amount of wireless transmission data and short sensor network lifetime results the inefficiency of odor monitoring.This thesis based on compressed sensing theory,studied the sampling,quantization and reconstruction process of toxic odor signals.It provides more advanced data acquisition technology and better processing algorithms for toxic odor leakage monitoring.And enrich and develop the monitoring technology of toxic odor leakage.The main work of this thesis as follows:Firstly,a block sparse model is proposed for sparse representation of odor signals.Studies have shown that the block sparse probability model can be used to express the intra-correlation of signals.Based on that,a block sparse multi-bit Bayesian compressed sensing reconstruction algorithm is proposed.The algorithm exploits the variational Bayesian inference method to reconstruct the signal,and estimates the parameters through the expectation maximization algorithm.Simulation results show that the proposed algorithm has good reconstruction performance for block sparse signals.Secondly,based on one-bit compressed sensing theory,the thesis proposed a block sparse one-bit compressed sensing reconstruction algorithm.The algorithm builds the Gaussian model for measurement noise,combined with the block sparse model of original signal to estimate the original signal through the variational Bayesian inference method.Simulation results show that the proposed algorithm can not only reconstruct the block sparse signal accurately,but also have robustness to noise.Finally,the validity and superiority of the proposed algorithms are verified by the real environment odor leakage monitoring experiment.The expected toxic odor leak monitoring system can not only be widely applied to monitoring of air quality,warning and emergency protection of sudden toxic odor leakage accident in the places of petrochemical industry and the storage of dangerous substances,but also be kept a large demand and has huge market potential for anti-terrorist of police,fire rescue and narcotics control of border areas.