Leveraging Machine Learning Techniques in Enhancing Recognition of Emotion in Speech

Ghanisht Aggarwal,Yash Khanna

DOI: https://doi.org/10.2139/ssrn.3842556

2021-01-01

SSRN Electronic Journal

Abstract:Emotions can play an important role in determining how we think and behave. Emotions compel us to take decisions be it small or big. In order to understand emotions, it is paramount that we understand the critical expressive component. While interacting with people, it is cardinal to provide clues in the form of emotions to interpret and react accordingly. In this work, to tackle the ambiguity of speech, we have adopted an engineering technique based on speech emotion recognition. Formalizing our concern as a multi-class classification model, we compare the performances of different machine learning models by extracting numerable artisanal features of the audio signal and employing them to train six conventional machine learning models. For the various experiment settings in which we tested our models, we document accuracy, f-score, accuracy and recall. We are able to achieve at par performances form Gradient boosting and Random Forest classifiers. Ultimately, we have shown that simpler machine learning based models trained over a few hand-crafted features are able to achieve performances that may be analogous to the current deep learning based state-of-the-art methods.