基于稀疏贝叶斯学习的码元速率估计

现有的相位编码信号码元速率估计方法在样本点足够多的情况下才能准确估计出参数，且算法复杂度高。针对此问题，该文详细分析了BPSK信号的结构特征，并以此为先验信息对其循环自相关(CA)向量进行压缩采样，降低了传统贝叶斯复数处理方法的维度。利用压缩传感中离散傅里叶变换矩阵的奇偶性，分解传感矩阵为正弦和余弦变换，分别将CA向量的实虚部转换到对应变换域测量，根据复数信号实虚部具有相同支撑集这一特点，采用多任务稀疏贝叶斯重构时延积向量的单边谱分量，从而估计出码元频率。理论分析和仿真结果表明，相较于其它基于稀疏贝叶斯学习的参数估计算法，所提方法在测量数量较少的情况下也能准确估计出循环频率，且算法实时性显著提高。

Symbol Rate Estimation Based on Sparse Bayesian Learning

Existing methods for symbol rate estimation of phase coded signals require amounts of sensing data, and are of high computational complexity. This paper analyzes the structure characteristics of BPSK signals, which are employed as the prior information for signal compressing and dimensionality reduction. The sensing matrix can be split into sine and cosine component, combined with the Fourier transform parity. According to the fact that the real and imaginary components of a complex value share the same support set, the symbol rate estimation can be obtained, using unilateral spectral of the delay-product vector reconstructed by multi-task Bayesian compressive sensing. Theoretical analysis and simulation results show that compared with other parameter estimation algorithms, the proposed method can reduce the measurements and significantly improve the real-time ability, while keeping the high reconstruction accuracy.