Emotion Classification from Speech using Prosodic and Spectral Attributes

LUCIANA FERRER; AGUSTIN GRAVANO; LARA GAUDER; PABLO RIERA

Ciudad Autónoma de Buenos Aires

Congreso; Machine Learning Summer School; 2018

Speech emotion detection is the task of recognizing the emotion of a speaker based only on their speech. The poster describes our work on this task using publicly available datasets. Prosodic (such as energy, voice quality or fundamental frequency) and spectral attributes are extracted from the audios and modeled using support vector machines, decision tree and different deep neural networks (DNN) architectures.