This will detect the emotions in your speech based on the MFCC, Delta, Delta-Delta & Prosodic Features (Fundamental Frequency) present in you voice. It makes use of GMM for classification. Try making models for basic emotions like happy, neutral, anger and sad. The emotion detection is text independent and speaker independent. But you need to train enough database in order to be speaker independent.
You need following things in order to make use of this repository.
-
Octave (For plotting data.) (You can ignore this if you don't want to plot features. But you need to edit source file.)
-
em4gmm You need to download em4gmm from https://github.com/juandavm/em4gmm
make
make install -
Opemsmile You need to download opensmile from http://www.audeering.com/research/opensmile
Install it using
sh buildStandalone.sh -p /usr/bin -
You may also need to download libsound using
sudo apt-get install libasound2-dev