Title :
On consistency of multi-innovation extended stochastic gradient algorithms with colored noises
Author :
Yu, Li ; Ding, Feng ; Liu, Peter X.
Author_Institution :
Control Sci. & Eng. Res. Center, Jiangnan Univ., Wuxi
Abstract :
This paper applies the multi-innovation identification theory to study the parameter estimation problem for CARMA models, and presents the multi-innovation extended stochastic gradient algorithm by expanding the scalar innovation to an innovation vector, and analyzes and proves the convergence properties of the algorithms involved. The simulation results show that the proposed algorithms are effective.
Keywords :
autoregressive moving average processes; filtering theory; gradient methods; parameter estimation; signal denoising; stochastic processes; vectors; CARMA models; MIESG algorithm; colored noises; innovation vector; multiinnovation extended stochastic gradient algorithms; multiinnovation identification theory; parameter estimation problem; Algorithm design and analysis; Colored noise; Convergence; Filtering algorithms; Linear regression; Parameter estimation; Stochastic resonance; Stochastic systems; Technological innovation; Vectors; Parameter estimation; convergence properties; martingale convergence theorem; multi-innovation identification; stochastic gradient;
Conference_Titel :
Instrumentation and Measurement Technology Conference Proceedings, 2008. IMTC 2008. IEEE
Conference_Location :
Victoria, BC
Print_ISBN :
978-1-4244-1540-3
Electronic_ISBN :
1091-5281
DOI :
10.1109/IMTC.2008.4547317
