A novel subband forecast method for nonlinear time series using wavelet transform

In this paper, a new method is proposed to implement subband forecast within the nonlinear noisy time series based on abstracting and reconstruction of the sign al's main components and adaptive Volterra filter theory.By considering noise's wavelet transform characteristic,the main component of noise signal is abstracte d by using the wavelet package decomposition in an appropriate scale and the ma ximum module reconstruction algorithm,then the forecast components are brought f rom adaptive Volterra forecast filter to reconstruction the final signal.This m ethod improves the traditional blindness in selecting scale in wavelet decomposi ng denoise,avoids the shortage of antinoise capability of Volterra series model used singly.The simulated results show that it is a practicable and effective me thod for nonlinear noise signal.