Smart Response System using Speech Emotion

Abstract

Speech recognition is becoming pervasive in many areas like smart-assistants, navigation systems and home automation to name a few. To build a truly smart speech recognition system, not only the content but also the emotion embedded in the content should be extracted. Generating responses according to human’s speech emotion is challenging as speech involves dealing with many ambiguous features. In this paper we propose to develop a system, which uses signal features like pitch, energy and Mel frequency cepstral coefficients, then uses Gaussian mixture models to detect human emotion which will further be used to generate output using extreme learning machine.