Mood Based Music Recommendation for a Mall using Real-time Image

Abstract

Human beings have the natural ability to guess the mood of a person just through their facial expression. This ability is highly valuable to be learnt by a computing device, like computers or robots. Music affects the human emotional core and their memories very deeply, thereby affecting our mood. Similarly, the background music played in malls affects a shopper’s behaviour. Customers who listen to music they like, are more likely to have a positive experience. Time becomes more pleasant even if it is spent waiting in line or waiting to speak to a customer service associate. The proposed system is built to give customers a better and more satisfactory experience, which would make them stay longer and purchase more. It also helps the workers to be in a better mood. All leading to an increase in sales in the mall. The system uses the latest crowd image and finds the 2 most common facial moods using a CNN. In the backend, the songs present in the database are classified based on audio and lyrical features. Finally, a playlist of songs is recommended based on a percentage mapping of moods found in facial features.