Abstract:
Speech Emotion Recognition (SER) has been an active area of research ever since the need for smooth and natural Human-Computer Interaction (HCI) came into play. This thesis aims to develop an SER system based on an amalgamation of Tensor Factorization and Neural Network-based learning to mitigate several issues while using contemporary deep learning architectures. This, in turn, is helpful towards recognizing the mental health issues such as depression, anxiety, etc., from speech signals as it is shown in the literature that mental health and emotions are highly correlated. As such,
this thesis tries to provide techniques to incorporate emotional information to assess mental health conditions from speech signals, thereby helping the psychologists assign a depression score to patients based on their experience and machine-generated score, thereby mitigating any human bias which might creep in human-only situations.
