Improvement And Implementation Of Unlicensed Spectrum Sensing Algorithm For LTE-U System

With the development of wireless communication technology,people have begun to pursue higher-speed and more reliable communication methods.These requires more spectrum for data transmission in mobile communications,which contradict the non-renewal nature of the inherent spectrum resources.Therefore,developing the existing Long Term Evulution(LTE)to the unlicensed spectrum(LTE Unlicensed,LTE-U)has become a simple and effective solution for spreading spectrum resources.At present,the unlicensed spectrum that close to 5GHz is widely used by wireless local area network(WLAN)and radar signals.In the existing LTE-U standard,user terminals need to follow different avoidance criteria when detecting different signals.Accordingly,the spectrum detection algorithm is required to have a signal identification function and low complexity.This thesis analyzes three traditional spectrum sensing algorithms in cognitive radio,and as a result only cyclic stationary detection can accomplish the signal recognition function within the LTE-U system in the practical applications.However,the traditional cyclic stationary detection algorithm has extremely high computational complexity,cannot meet the hardware requirements of the terminal chip.This thesis further explores three different implementation schemes of cyclic stationary detection algorithm,and proposes an optimized FFT accumulation algorithm(FAM)based on one single surface.Some appropriate constraints need to be used when constructing a cyclic spectrum.Instead of constructing a full cyclic spectra,a certain aspect of cyclic spectra is constructed for cyclic stationary feature detection.By using performance simulation and complexity analysis,compared with the traditional FAM algorithm,the optimization algorithm has a lower detection performance when SNR is below-20 d B,and the same detection probability can be achieved at-18 d B.However,the optimized FAM algorithm can reduce the computational complexity by more than 10 times compared with the traditional FAM algorithm,and the optimization effect becomes more obvious as the sampling frequency increases.Based on the Maotu DSP,an optimized FAM algorithm is implemented in this thesis.According to the process of optimizing the FAM algorithm,the scheme design of two key modules(the DSP row-column transformation and the peak search)has been completed,which further improve the degree of parallelism of the optimization algorithm proposed in this thesis.Furthermore,the comparison of the optimized FAM algorithm performance between floating-point and fixed-point confirms that the performance difference is only0.2d B,proving the feasibility of the fixed-point algorithm.