Emotion Analyzer From Youtube Audio

This repository allows you to retrieve emotion in two classes: Arousal and Valence. The youtube audio is downloaded and emotion analysis is done on full spectrum, foreground and background.

Requirements

To work with this repository, you need to have following dependencies installed on python 2.7

pip install youtube_dl

pip install librosa

pip install pyAudioAnalysis

once the pyAudioAnalysis is installed, run the following command:

python audioAnalysis.py trainRegression -i data/speechEmotion/ --method svm -o data/svmSpeechEmotion

Testing using Jupyter NoteBook

Make sure to make changes in model path and audio slice dir in jupyter notebook before running. https://github.com/babupriyavrat/EmotionAnalyzerFromYoutubeAudio/blob/master/Emotion%20Detection%20from%20Youtube%20Audio.ipynb

Testing using test.py

Change the url in test.py and run the code.

Parallel processing

In order to speed up wav file generation, you can run wavechunk_generator.py based on the number of CPUs you want to assign to the wav file generation. Change the number of threads.