Research On Sparse Recovery Algorithm And Its Application In Communication System

Massive,Inexact,and Unlabeled information is constantly generated and transmitted in varies respects of the social system.How to convey determined signal both quickly and accurately? How to reconstruct and identify undetermined signal accurately? How to classify and track unlabeled data accurately? How to identify the unknown system accurately? As for the above problem,which can be converted into linear equations directly or indirectly.It is crucial to mine potential prior features.Robust and Efficient Sparse Recovery algorithms make contributions to solving this problem.However,existing algorithms need to be further improved in terms of reconstruction accuracy.In order to overcome the deficiencies of the traditional ones,a novel SL0DL2 algorithm utilized both SL0DL2 constraint and proximal operator is proposed in this thesis.The system is based on the graduate sparse framework of the SL0 algorithm.Our scheme achieves exciting performance via sparse constraint and optimization strategy.Experiment results indicate the proposed SL0DL2 algorithm is superior to the typical sparse recovery algorithm in all key performance indicators.High-speed and reliable communication is the foundation for our daily life and social development.Unfortunately,the actual communication environment is not ideal.Complex noise degrades the quality of communication,especially the existence of impulse noise.The traditional impulse noise mitigation scheme does not fully exploit the prior characteristics of the signal,which results in in unsatisfactory consequences.In order to achieve a better performance.By making full of use both the sparse prior to the impulse noise and null tone feature,the impulse noise mitigation system based on sparse recovery algorithm is built in the thesis.The scheme is operated in three typical noise impulse noise scenarios.Experiments show that the sparse recovery algorithm can be used in impulse noise mitigation effectively,espically the SL0DL2 algorithm outperform than other ones.Indoor localization based on Receive Signal Strength Indicator(RSSI)is widely used due to its low infrastructure cost.Unfortunately,RSSI is always fluctuating because of multipath effect,human movement and environmental change,which makes traditional methods get unexpected results.What's more,state-of-the-art clustering algorithm divides offline radio map into the different non-overlapping group while ignores the fact that different groups should share the same members.In this paper,a novel Robust Sparse Overlapping Group Lasso(RSOGL)algorithm is proposed for indoor localization.The scheme first utilizes similarity between the offline Reference Point(RP)and online Test Point(TP)to obtain overlapping group via Fuzzy C-means(FCM)clustering,and then uses Fo Glasso algorithm to reconstruct TP's fingerprint.Our RSOGL system has been operated in a real environment.Experimental results demonstrate that proposed system outperforms traditional fingerprinting methods.