VoiceSense: Designing A Robust Emotion Detection System

Detecting emotions from speech is an excitingarea of research that can improve how we interact with computers and understand mental health. In our study, we explore how to build a system that can understand different emotions using speech and psychological information. We gather data from various sources like TESS, Crema-D, RAVDESS, SAVEE, and our own custom dataset to create a diverse collection of emotions. To build our system, we use different types of neural networks like CNNs, RNNs, and transformers. After testing our system extensively, we find that it can accurately detect emotions about 77% of the time, which is pretty good. We also look into how our system learns about emotions, which can help us better understand how humans process emotions. Our work not only improves emotion detection technology but also has important implications for making computers more empathetic in how they interact with people.