期刊名称:International Journal of Innovative Research in Computer and Communication Engineering
印刷版ISSN:2320-9798
电子版ISSN:2320-9801
出版年度:2015
卷号:3
期号:4
DOI:10.15680/ijircce.2015.0304175
出版社:S&S Publications
摘要:In this paper of a robust metric learning Approach on facial expressions using texture feature and KNNBased classifications, we basically emphasised on two factors in this field. The 1st and important thing is inherentsubtlety, the appearance features and geometric features of spontaneous expressions basically the overlap with eachother, so as to make it difficult for classifiers to find the effective separated boundaries. And the 2nd thing on which weare emphasising is, in all the training set it basically comes with a dubious class labels which can create an obstacle inrecognization performance if no measure should be taken. In this paper we are implementing a new method calledspontaneous expression recognization process, which is based on roboust metric learning so as to sought out with thetwo important issues in this paper. The most important requirement here is to increase the discrimination level in thedifferent facial expressions. We got to know a new metric space in which a more number of chances are thereof sameclass is possessed by the spatially close data points. If we emphasis more in this , So to characterize all annotationreliability, we can define by specificity and sensitivity one annotator, instead of using the noisy level directly for metriclearning techniques. The various comparative experiments can show us that our experiments have success percentageas compare to others in spontaneous facial expression recognization field and can be changed to recognize various otherexpressions.