Our new study “Unveiling Emotions in Speech: A Multi-Dimensional Transformer Approach” explores using transformer models to classify emotions in human speech. By analyzing acoustic features like arousal and valence, we trained multiple transformer architectures. 

Our experiments found a 3-transformer design achieved up to 90% accuracy in detecting emotions like happiness, sadness, and anger in speech samples. This research demonstrates the potential of AI to perceive emotions for more natural human-machine interactions.