A Novel Learning Algorithm for Mixture Gaussian Models
Zhu Xiao-zheng · Aeronautical Computing Technique · 2006
Mixture Gaussian model is one of background subtraction methods.An effective and adaptive learning method to update parameters and increase learning rate is proposed in this paper.Comparing to previous work there are two innovations in the new algorithm.First,the parameters can be computed online through the recursive equations according to maximum likelihood rules.Second,forgetting factors and learning rate factors are redefined and their general and simple formulations are obtained by analyzing their practical functions. The new algorithm is applied to simulating data and actual video.The results show that the proposed learning algorithm excels the formers both in converging rate and accuracy.