Disruptive situation detection on public transport through speech emotion recognition

Eleonora Mancini, Andrea Galassi, Federico Ruggeri, Paolo Torroni

Abstract: Highlights•We address an under-investigated, yet socially significant problem: Disruptive Situation Detection in Public Transport.•In agreement with cognitive science theories, we approach the detection task as a Speech Emotion Recognition task.•We develop a model characterized by speaker/gender independence and robustness to noise and presence of multiple voices.•We introduce a data-augmentation process to obtain realistic domain-specific recordings.