ABSTRACT:- One Now a days to human-machine interaction is estimating the speaker’s emotion is a challenge. The need is more accurate information about consumer choices increasing interest in high-level analysis of online media perspective. Usually in emotion classification, researchers consider the acoustic features alone. For strong emotions like anger and surprise, the acoustic features pitch and energy are both high. In such cases, it is very difficult to predict the emotions correctly using acoustic features alone. But, if we classify speech solely on its textual component, The uniqueness in this approach is the generation of text sentiments, audio sentiments and blend them to obtain better accuracy. In this paper, I have proposed an approach for emotion recognition based on both speech and media content. Most of the existing approaches to sentiment analysis focus on audio and text sentiment. This novel approach is the generation of text sentiments, audio sentiments and blend them to obtain better accuracy.

Keywords:- Sentiments, Features, Natural Language Programming, Hybrid, Accuracy.