Emotion Recognition From Audio Signal

This is a set of soultions of Emotion Recognition From Audio Signal from Surrey Audio-Visual Expressed Emotion (SAVEE) Database using OpenSmile, Principal Components Analysis and set of classifiers from Scikit-learn library.

The SAVEE database was recorded from four native English male speakers (identified as DC, JE, JK, KL), postgraduate students and researchers at the University of Surrey.The database consists of recordings in 7 different emotions, 480 British English utterances in total.

OpenSmile is used to extract features from .wav file, it generate 1582 feature. After that, I applied PCA (Principal Components Analysis) for dimentionality reduction. Finally, I used different algorithms for classification.

The different emotions are:

Anger
Disgust
Fear
Happiness
Sadness
Surprise
Neutral

Results

Algorithm	Model Performance
Multi Layer Perceptron Classifier	80.20%
Logistic Regression + lbfgs Solver	76.04%
Linear Discriminant Analysis	76.04%
Support Vector Classifier	75%
Logistic Regression	75%
K-Neighbors Classifier	66.66%
Gaussian Naive Bayes	64.58%
Decision Tree Classifier	54.16%
Ada Boost Classifier	41.66%
Random Forest Classifier	32.29%
Discriminant Analysis	28.12%
Gaussian Process Classifier	17.70%

Requirements

Python 2.7 and up

Installation

The followoing are the prerequiste Python modules that needs to be installed to execute main.py:

sudo pip install pandas 
sudo pip install -U scikit-learn

Downloads

Clone the repository using the below mentioned command and execute the Python program.

git clone https://github.com/Sehaba95/Emotions-recognition-from-audio-signal.git
cd Emotions-recognition-from-audio-signal/Classifiers
python MLPClassifier.py

Authors

Sehaba Amine