Human Affect Recognition Using Deep Learning Approaches

An important advance towards more characteristic human-computer interaction is the automatic detection of human impact. Present-day developments entail the integration of physiological and audiovisual sensors to identify emotions in real-world situations, posing a challenge to standard machine learning algorithms. A thorough assessment of creative writing on human affect recognition from 2010 to 2017 is done in this research, with a centre on deep neural network-based approaches. A distinct trend toward the promotion of deep learning (DL) methods is visible in the classification of 950 studies based on whether they used shallow or deep learning architectures. An extensive assessment of deep neural network appeal in this sector is possible through a more concentrated investigation of a subset consisting of 233 papers that use them. The development of spatial, temporal, and joint feature representations for multi- modal sensor data requires DL. To demonstrate the significant advancements made in this area, the study presents estimable state-of-the-art architectures. For academics working on similar applications, the outcome is identified, and the significance of frameworks in the field of human affect is recognized.