The main objective of this project is to build a model for Music Emotion Recognition(MER) using both components of music which is lyrics and audio by comparing various preprocessing methods, feature engineering technique and applying various classification models to 16k lyrics and audio data.
Objective is to classify the music into 4 classes of emotion based on Russell’s V-A model: sad, angry, happy and calm.
The accuracy by integrating the result from lyrics and audio data was 90.14%
Language: Python
Description about files uploaded:
Presentation - Contains a ppt about the project
Lyrics - Contains the code to web scrape lyrics data and preprocess the lyrics
Lyrics Classification - Contains the code to classify emotions using lyrics features
Dataset - Contains the code to web scrape audio data
Feature Extraction - Contains the code to extract features from audio dataset using various preprocessing and feature engineering methods
Emotion Recognition - Contains the code to classify emotions based on features using various classification models and result integration
train.xlsx - Contains all the features extracted from trainD audio dataset in excel file
test.xlsx - Contains all the features extracted from test audio dataset in excel file
validation.xlsx - Contains all the features extracted from validation audio dataset in excel file