时频信号的非参数加窗稀疏协方差迭代分析法

针对多分量非平稳时频信号的分析，提出了一种非参数加窗稀疏协方差迭代分析（Sparse Iterative Covariance-based Estimation，SPICE）法。该方法通过引入局部化的窗函数，并在窗内对信号建立非参数化的时频模型，以获取非平稳信号的局部时频特性。分析表明：给出的非参数时频模型，可通过加权最小二乘（Weighted Least Square，WLS）进行求解。当将WLS的加权矩阵的构建问题转换为广义噪声协方差矩阵时，提出的模型也可借助于加窗SPICE方法来进行求解。分析和仿真均表明：与传统方法相比，提出方法具有噪声抑制能力强、能量集中度高和对于载频差异较小信号的分离效果佳等优点。

A Nonparametric Windowed Sparse Iterative Covariance-based Estimation Approach to Time-Frequency Signals

A nonparametric windowed Sparse Iterative Covariance-based Estimation (SPICE) approach for time-frequency signals analysis is proposed. By introducing a localized window function, a nonparametric time-frequency model in a window function is established for acquiring the features of the non-stationary multi-components time-frequency signals. Then, it is shown that the given nonparametric time-frequency model can be solved by the Weighted Least Square (WLS) method. When the WLS weighted matrix is replaced by a generalized noise covariance matrix, it is found that the given model could also be solved by a SPICE windowed method. Both the analysis and simulation result have shown that the proposed method is superior to the traditional one for its better noise suppression effect, high energy concentration, and an outstanding separation effect on the signal with smaller carrier frequency difference.