基于快速变分稀疏贝叶斯学习的频谱感知与定位

Spectrum Sensing and Location Based on Fast Variational Sparse Bayesian Learning

针对稀疏贝叶斯压缩感知算法存在复杂度高、收敛速度慢等缺陷,提出了一种快速变分稀疏贝叶斯学习的频谱检测与定位算法.该算法在原始问题求解过程中增加了辅助变量,消除了原问题模型中未知变量之间耦合度高的问题.并依据稀疏参数的收敛情况,自适应删除不收敛稀疏参数对应的基函数,从而进一步加快了算法的收敛速度.实验结果表明:该算法在收敛速度和频谱检测精度上有显著的改善.

Based upon the fact that sparse Bayesian compressed sensing algorithm has the defects of high complexity and slow convergence speed , a spectrum sensing and location algorithm based on fast variational sparse Bayesian learning is proposed.The algorithm adds some auxiliary variable in the process of solving original problem , which eliminates the high coupling coefficient between the unknown variables in the original model .At the meantime, the algorithm can adaptively delete the basic functions corresponding to un-convergence sparse parameters according to the converging conditions of the sparse parameters , thus leading to the effect that the velocity of convergence is further accelerated .The experimental results show that the algorithm significantly improves the accuracy and speed of sensing .