Emotion recognition and speaker verification using machine learning

The aim of the work is to develop machine learning algorithms that can verify the identity of a speaker and infer their emotional state. Emotion recognition utilizes a person’s voice characteristics to infer their emotional state and has been used for several applications including quality metrics on call centres, determining patients feeling etc. In a different vein speaker verification utilizes voice extracted features to properly identify individuals and has been used for several applications including security systems, authenticating speakers, etc. State-of-the-art approaches in both fields are achieved through deep learning techniques. Deep learning models trained in an end-to-end fashion do not require the guidance from experts to achieve satisfactory performance in their tasks. A rich dataset of examples depicting the desired results is enough to allow the models to generalize well to unobserved scenarios. 

Basic knowledge of machine learning process and python programming language.

Bridging technologies in a web application is becoming the norm in companies, hence the internship objective will be to develop a web application that can verify speakers and evaluate their emotional state. A speech-oriented web application will be developed utilizing the Django web framework. Initially, a research on the current advancements in emotion recognition and speaker verification will be performed. Based on these findings a choice of the deep learning models will be made and implemented using the Tensorflow library. The models will be trained with publicly available datasets and embedded in a web application using the Django web framework. Finally, a feasibility study will be followed on the computational complexity of the software and their readiness to be deployed for real-time scenarios. The final product will be tested by in-house volunteers and their feedback will be collected. The objective is not only to develop an innovative web application but also to experience a full cycle of software development.