Abstract:
"Nowadays people are more competitive with their lifestyle. Because of that a lot of them suffer
from mental stress. Because of this problem many people failed in their career, some of them are
trying to lose their life. Studies proved that music could relieve stress and music is the best
medicine for many diseases. But when people select music, they pick a random song and listen to
it. But it’s never going to help to reduce stress. Sometimes it can increase stress.
By considering the above details music recommendation system based on expressions introduced.
It captures the users’ facial expressions through their devices and provides necessary songs
matching their facial expressions.
To implement the model CNN model is used and as the dataset FER 2013 dataset utilized. In FER
2013 dataset there were 25102 train images present and 6236 test images present. Out of these
data, the model achieved a quite high accuracy but it’s not enough. The researchers suggest that a
more precise data set and the addition of more image data could be used to identify complicated
emotion categories.
The output of the system facial emotions is not very accurate when testing the facial expressions."
