Paper Info
Title
Robust estimation of an AR multi-channel model by using t-distribution assumption
Abstract
In this paper, we propose a new error criteria for determining the optimal multi-channel model system. The error criteria is based on assuming that the probability density function of the resulted error signal is t-distributed with α degrees of freedom. A small weighting factor is assigned for large amplitude signal portion parts and large weighting factor is used for small amplitude signal portion sections. By doing so, the effect of large amplitude signal portions to the estimated system parameter is reduced. The simulation results show that the average of the obtained parameter by using small degree of freedom α t-distribution assumption is closer to the ideal parameter than that when the conventional Gaussian assumption is applied. Furthermore, the standard deviation of the estimation result by applying small α t assumption is smaller than that when α = ∞ is utilized.
Year
Venue
Keywords
2000
EUSIPCO
simulation,robustness,estimation,probability density function
Field
DocType
ISBN
Degrees of freedom (statistics),Weighting,Control theory,Mathematical analysis,Robustness (computer science),Gaussian,Probability density function,Amplitude,Standard deviation,Mathematics,T distribution
Conference
978-952-1504-43-3
Citations 
PageRank 
References 
0
0.34
0
Authors
2
Name
Order
Citations
PageRank
Junibakti Sanubari142.23
Keiichi Tokuda23016250.00