Differential Evolution to Optimize Hidden Markov Models Training: Application to Facial Expression Recognition

摘要: The base system in this paper uses Hidden Markov Models (HMMs) to model dynamic relationships among facial features behavior interpretation and understanding field. input of HMMs is a new set derived from geometrical distances obtained detected automatically tracked points. Numerical data representation which the form multi-time series transformed symbolic order reduce dimensionality, extract most pertinent information give meaningful humans. main problem use that training generally trapped local minima, so we used Differential Evolution (DE) algorithm offer more diversity limit as much possible occurrence stagnation. For reason, proposes enhance HMM learning abilities by DE an optimization tool, instead classical Baum Welch algorithm. Obtained results are compared against traditional approach significant improvements have been obtained.