Study on autoregressive trispectrum slices

A fourth-order cumulant for trispectrum is presented based on autoregressive (AR) spectrum estimates to replace discrete Fourier transformation (DFT) in this paper. Three methods-the AR trispectrum, AR bispectrum and AR power spectrum methods are studied and compared with each other using some corresponding slices from some charts in detail. Experimental results show that the AR trispectrum has much better performances and better distinguishability than the AR bispectrum and AR power spectrum. The experimental and theoretical results show that the autoregressive trispectrum model is helpful to establish the dynamic model and more accurate to analyze the non-linear and non-Gaussian signals. The measurement and analysis process is based on virtual instruments and automatically controlled by computer.