Investigation of Emotion Instances and Class Analysis from Physiological Sensors by Unsupervised Hybrid EMDeep Model

International audiencePattern of emotion identification is one of the improvised research application regarding facial expression as major concern, in those cases, conventional facial expressions for patterns identification. The present model is based on signal collected from phys-iological sensors followed by consecutive deployment of unsupervised machine learning model. The proposed model is unsu-pervised in following aspects: firstly, it introduces Expectation Maximization problem with respect to unknown emotion labels to be derived from the measures. Correlation of physiological signal and individual emotion labels can be identified. This follows a considerable emotion classification method. However, the output of EM model doesn’t ensure the correct identification of emo-tion class, if any. We introduce Support Vector Regression (SVR) as output module of this model. Hence, we try to forecast the probable classes of emotion after investigating the ranges of values and appropriate standard threshold values of physiological signal with respect to respective emotion class e.g. angry, frustration and joy. This should be noted that, the proposed model doesn’t envisage facial expression analysis. However, after successful implementation of Gaussian behaviors of mixed physio-logical signal, we can enhance the accuracy of identification. Significant emotional context exists in output with more precise results of emotion identification phases