Hanran: Harmonium Note Recogniton Using Artificial Neural Networks

Abstract

Lately, in recent Decades, Music has become popular among youth. Not only for listening to music, but learning music or musical instruments have also become a popular choice of most of the population globally. Playing musical instrument has also become new trend for youth. There is a need of technological assistance for determining the exact string/note/chord played over the musical instrument. The notes are the primary or basic unit of music (musical instruments). So for a person to play the instrument well, person must know the exact notes to be played. The proposed system emphasizes on designing an ANN (Artificial Neural Network) for determining the played notes from the input music audio. ANN extracts the features from the input audio file and based on the features, it predicts the played note. Prior to ANN, manually features were extracted using MFCC or CQT but simple ML Based approach failed to gain maximum accuracy. So Neural Network Based Approach is proposed. Some feature combinations such as Onset Detection, Chromagram, CQT, Speech Frequency gives around 88%. This system accomplishes the promising results on the self-generated/synthetic database.