An integrated approach to emotion recognition and gender classification

The human-computer communication adds to worldwide access to training, information upgrade and the conveyance of value learning and instructions. Furthermore, the changing paradigm of machine perceiving the human emotions effectively is a technological advancement for the learning community. The automatic recognition of emotions reflected from speech and facial expressions for making the machine to understand the human verbal and non-verbal emotions is coined as Emotion Recognition. Emotion recognition systems, regardless, are not totally seen as subject independent dynamic features, so they are not sufficiently vigorous for constant acknowledgment assignments with subject assortment (human face), head movement, illumination change, speech, noise and so on when thought about exclusively. Hence, it is proposed to fuse the features of facial expressions and speech. The present framework utilizes the Speech (Mel Frequency Cepstral Coefficients) features and Facial (Maximally Stable Extremal Regions) features to predict the emotions of a person through a systematic and scientific study. Specifically, when combining MSER with MFCC, the recognition rates can be further improved by 2 to 3% on Indian Face Database and Berlin Speech Database. (C) 2019 Elsevier Inc. All rights reserved.