AN INTEGRATED APPROACH FOR MUSICAL INSTRUMENT RECOGNITION USING MACHINE LEARNING

Abstract

One of the most significant components of life is music. Social media platforms are used by a large number of musicians to share their songs. Finding a specific one depending on the performer, genre, song, instrument, and so forth is challenging and time-consuming. Prior efforts primarily focused on classifying distinct instruments from different families, such brass, woodwind, string, percussion, and so forth. For Musical Instrument identification, audio signals are often converted into Fourier domains and then examined. There are methods for analyzing the audio signal utilizing several audio descriptors. It is challenging to select the right audio descriptors from a pool of hundreds in order to use them as features of a musical instrument recognition. This research paper,  propose an Integrated approach called Integrated Entropy with audio features for enhancing classification accuracy in order to identify musical instruments. In this research work, audio features have taken into consideration for Temporal aspect – Zero Crossing Rate(ZCR), Spectral aspect- Spectral rolloff(Sr), Spectral spread(Ss), Spectral centroid(Sc), perceptual aspect -Roughness(Rg), Brightness(Br), Flatness(Fl). In the first experiment audio features have been independently (AFI) given to classifier whereas in Integrated approach, audio features are integrated with Entropy of audio. The highest accuracy is obtained   This approach is very useful to improve the classification accuracy of Machine learning model.