OCCLUDED FACIAL EXPRESSION RECOGNITION USING ALEXNET
In artificial intelligence and machine learning fields, facial emotion recognition (FER) is receiving an attention to research as it is an essential commercial and academic potential. By using facial images, this analysis is performed while multiple sensors can use for processing the FER. In interpersonal communication, one of the key channels is that the visual expressions. In the field of FER, a details analysis of research is performed over the past few decades in this paper. The conventional approaches of FER are listed out in addition to the describing representative categories of FER systems and their key algorithms. By using deep networks, the approaches of deep-learning-based FER are presented that allow the “end-to-end” learning. An approach of up-to-date hybrid deep-learning is also focused in this research and it integrates an individual frame’s spatial features with the convolution neural network (CNN) for consecutive frames’ temporal features. Consequently, the publicly available assessment metrics with a brief analysis and defining of comparing the benchmark results have been provided for a quantitative comparison of FER studies. The existing work carried out was using basic CNN variant which couldn’t produce efficient results for larger datasets. This paper proposes a facial recognition system using ALEXNET which produced better results than the existing basic CNN and GoogleNet.