基于分数阶Fourier变换的PD时频特征提取

Time Frequency Feature Extraction of Parallel Data Based on Fractional Fourier Transform

针对单一信道信号时频特征提取时无法过滤并行数据,导致交叉信号抗干扰能力较差、信号时域特征提取精度较低等问题,提出分数阶Fourier变换的并行数据时频特征提取方法。采用分数阶傅里叶变换重构并行数据信号,利用窗函数过去短时分数阶傅里叶变换信号;根据架构并行数据一维向量组合,调整参数完成数据压缩;通过非负矩阵经初始化与迭代处理,实现目标函数最小化;在获取的频域基向量以及相应时域位置向量后,提取尖锐度、信息熵、稀疏度等时频特征参量,进行归一化处理,完成并行数据时频特征提取。仿真结果表明,采用所提方法进行并行数据时频特征提取的精度较高,且交叉信号的抗干扰能力较强。

Currently, the problem that parallel data can’t be filtered when extracting time-frequency characteristics of single channel signals leads to low anti-interference ability of cross signals and low accuracy of time-domain feature extraction. Therefore, a method to extract the time-frequency feature of parallel data based on the fractional Fourier transform was proposed. First, the parallel data signal was reconstructed by the fractional Fourier transform. Second, the short-time fractional Fourier transform signal was processed by the window function. The combination of one-dimensional vectors of parallel data was used to adjust the parameters, and thus to complete data compression. Then, the objective function was minimized through the initialization and iteration of a nonnegative matrix. After obtaining the frequency-domain base vector and the time-domain vector, some time-frequency feature parameters such as sharpness, information entropy, and sparsity were extracted and normalized. Finally, the time-frequency feature extraction of parallel data was completed. Simulation results show that the proposed method has high accuracy for time-frequency feature extraction of parallel data, and the anti-interference ability of cross signals is better.