Abstract:
Determining the songs that the user want to hear is a challenging task, due to many reasons the result of the user playing songs goes to decline. The focus of our project is creating a system that can precisely anticipate the emotion of the user. the prediction is dependent on the input of voice/speech of the series of question and activation audio . In fact, sometime the user does not know what song they want to hear. They do not know whether they want to hear happy or sad music, Our model plays a vital role in to overcome this situation. We have developed a system that takes input using microphone store it in a .wav file and use it further to predict emotion. Long short term memory algorithm is used as it showed best accuracy and learn through it too, continuously improving the system accuracy. Using this, it will be possible to predict the emotion of the user efficiently. We have considered various feature extraction measures i.e. MFCC, zero error crossing rate and Root mean square. After prediction, user have the free hand to decide whether he wants to play music from local playlist created or on YouTube
