A novel ELM based adaptive Kalman filter tracking algorithm

• Published in: Neurocomputing (Amsterdam) Vol. 128; pp. 42 - 49
• Main Authors: Chi, Jian-Nan, Qian, Chenfei, Zhang, Pengyun, Xiao, Wendong, Xie, Lihua
• Format: Journal Article Conference Proceeding
• Language: English
• Published: Amsterdam Elsevier B.V 27.03.2014; Elsevier
• Subjects: Adaptive algorithms; Adjustment factor; Algorithms; Applied sciences; Artificial intelligence; Computer science; control theory; systems; Connectionism. Neural networks; Covariance; Covariance matrix; Detection, estimation, filtering, equalization, prediction; ELM; Exact sciences and technology; Information, signal and communications theory; Kalman filter target tracking; Kalman filters; Noise; Online correction; Pattern recognition. Digital image processing. Computational geometry; Signal and communications theory; Signal, noise; Target tracking; Telecommunications and information theory; Tracking; Adjustment factor; Kalman filter target tracking; Covariance matrix; ELM; Online correction; Target tracking; Statistical analysis; Tracking; Adaptive algorithm; Gaze; Extreme learning machine; Kalman filter; Adaptive filter; Pursuit eye movement; Adaptive method; Pupil; Digital filter; Innovation; Feedforward neural nets; Covariance; Residue; Robustness
• ISSN: 0925-2312; 1872-8286
• DOI: 10.1016/j.neucom.2013.03.052

In order to avoid the filter divergence problem in target tracking caused by the unknown or changing statistical characteristic of the noise in Kalman filter, a novel ELM based adaptive Kalman filter tracking algorithm is proposed in this paper. By learning the difference between the theoretical covariance and practical covariance of the innovation which is defined as measurement residue through ELM, the adaptive factor of the covariance matrix of the observation noise was obtained. Then the covariance matrix of the observation noise can be adjusted online according to the ELM learning data. Simulation results showed that the proposed algorithm can improve the estimation accuracy and the robustness of the Kalman filtering for target tracking. It is also applied in the gaze tracking system for pupil tracking and shows satisfactory results.