International Journal of Computer Applications (0975 – 8887) Volume 185 – No. 1, April 2023 7 Music Recommendation based on Facial Emotion Detection Sonika Malik Assistant Professor Department of Information Technology Maharaja Surajmal Institute of Technology, Delhi, India ABSTRACT Human face expressions directly express what’s going on inside a person. Detecting someone’s emotions is not a difficult task for a human but for a machine it can be difficult. Also when it comes to sear a song according to our mood, is very difficult and confusing task. So, keeping in mind here we have designed a model which can easily detect what’s going on inside a human and recommend him few songs according to his mood using facial expressions. On the basis of the prediction of emotions, the goal is to play the song that best fits the mood reflected by our expression. The major task in the paper involves the detection of human face, extract the features of face and detect emotion, predict the emotions of new face, and play song on the basis of that emotion. Keywords Facial Emotion Detection 1. INTRODUCTION Music has the ability to excite powerful emotional responses. Listening to music is an easy way to change mood or relieve stress. There a lot ofdifferent form of music which reflects different emotions / moods of a person? People tend to listen music according to their changing moods. But they have to select the music manually by going through playlist of songs and play song on the basis of their current emotional state whenever they need so. But to avoid the difficulty to select a song, some people might play any random song which may be not a fit to their current mood and may disappoint the user. The idea of recommending a song based on thecurrent mood of a user will not only relieves the stress and enlighten the mood of the user but at the same time will save their time and remove overhead of selecting a song from their playlist. People express their emotions through facial expressions. Humans make use of facial expression to express clearlywhat they want to say. With the help of recommendation system, it could help a user to decide the music one should listen therebyhelping the user to reduce his/her stress level. In other words, this feature lightens up the mood of the user by playing those songs that match the requirements of the user by capturing image of the user. Facial expression is the best form of expression analysis that is being known to human kind. People can conclude the emotions/thoughts of another person with the help of their facial expressions [1, 2]. The aim of this paper is to capture the image through webcam and display the emotion and play song according to the displayed emotion. The methodology used here is CNN (Convolutional Neural Network). A CNN is most commonly applied to analyzing visual imaginary. So CNN turns out to be the best way to figure out the tasks in computer vision. Working on human emotion detection using CNN is the most preferable way of doing it as the chances of error is very less here. In this article five-layered CNN is used. The project focuses on brighten the mood of the user by playing the song that match mood of the user. The facial expression is broadly categorized into seven types like happy, angry, sad, disgusted, fear, surprise and neutral [3]. This project also makes use of the C++ dlib library which increases the overall performance of the project [4]. On further expansions of this paper we can help study human face and form lie detector using facial expressions. Not only this but by adding more features to it, may help in us to achieve many things like detection of human behavior in suspicious circumstances, which may help in reduction of many bad elements of society and also in huge reduction of terrorism. The following are the contributions: • To implement the ideas of machine learning, deep learning and neural networks for entertainment. • To provide a platform with new features for music lovers and an interface between the music system and users. • To reduce the headache of searching music and provide a new entertainment for the users. 2. RELATED WORK With Emotion Detection and Characterization Using Facial Features [5], faces can be detected from any given image, their features (eyes and lips) extracted, and their emotions classified into six categories (happy, fear, anger, disgust, neutral, sadness). Using Grid Search, the training data is refined after it passes through several filters and processes. A classification report is then generated based on the testing data and its labels. The best results have been obtained by passing the training images through HOG followed by SVM characterization, resulting in an average precision of 85%. An approach to improving music recommendation systems is presented in a novel music recommendation system using deep learning [6]. Various platforms and domains could benefit from the proposed solution, such as YouTube (videos), Netflix (movies), and Amazon (shopping). As more variables are introduced, current systems become inefficient. As input to a deep learning classification model, Tunes Recommendation System (T-RECSYS) combines content-based and collaborative filtering. Based on Spotify Recsys Challenge data, the authors achieve precision scores of 88% at a balanced discrimination threshold. Researchers developed a model to identify a model that recognizes facial micro-expressions and recommends music based on mood in Research on Automatic Music Recommendation Algorithm Based on Facial Micro-