【摘要】 双线性时间-频率分布是非平稳信号处理的重要工具。但该方法运算量大,占内存多,无法实时实现。根据核函数的谱分解理论,把实离散核函数表示为实对称矩阵,利用对称矩阵的特征分解把离散时间双线性时间-频率变换表示为离散时间频谱图的加权和,用频谱图的部分和实现双线性变换的快速近似计算。数字仿真与齿轮箱故障信号处理结果表明,近似计算方法能够节能存贮空间,明显提高计算速度。更多还原

【Abstract】 Bilinear time-frequency transformation has become an important processing tool for non-stationary signal.But large amount of computation and occupying big memory make it hard to real-timely implement.According to the principle of kernel function spectral decomposition,the authors express the discrete real kernel as a symmetric matrix,and indicate the discrete bilinear time-frequency transformation as a weighted sum of discrete spectrograms with eigen-decomposition of symmetric matrices.Then the bilinear transform can be fast cornputed approximately with the partial sum of spectrograms.The numerical simulation and signal processing of gearbox early fault vibration show that the algorithm has a smaller memory cost and a faster speed than traditional ones.更多还原