基于前后向线性预测的时变参数估计方法研究

Time-varying Parameters Estimation Method Based on Forward and Backward Linear Prediction

在基时间函数展开理论基础上,通过同时引入非平稳信号的前向线性预测和后向线性预测来估计模型的时变系数,提出了一种新的时变参数建模方法,有效地克服了现有方法对瞬时频率估计的滞后问题和无法准确估计初始时刻值的困难;仿真试验证明,在相同阶次和基函数维数条件下,该方法对非平稳信号瞬时频率的估计效果要优于现有的方法。

Based on the time-varying basis function expansion theory, a new method combining forward and backward linear prediction of non stationary signals is presented to estimate TVAR parameters. Computer simulation shows that the method can solve the one-step delay problem and overcome the computing difficulties at initial time steps in estimating the instantaneous frequency. The results show us that this method is more effective than the present approaches in modeling a non-stationary signal under the same TVAR order and basis function dimension.