Optimum Feature Selection for Harmonium Note Identification Using ANN

• Published in: 2019 10th International Conference on Computing, Communication and Networking Technologies (ICCCNT) pp. 1 - 7
• Main Authors: Puri, Surekha B., Mahajan, Shrinivas P
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.07.2019
• Subjects: Acoustic Model; ANN; Artificial neural networks; Automatic Music Transcription; Correlation; Feature extraction; Filter banks; Indian Classical Music; Mel frequency cepstral coefficient; Microsoft Windows; Music
• DOI: 10.1109/ICCCNT45670.2019.8944560

The harmonium music notes exhibit multiple variations so the selection of the optimal features for the musical note identification is difficult. To improve identification accuracy, the selection of the important features is necessary. In this paper, an artificial neural network (ANN) based automatic harmonium musical note identification approach is proposed by considering different features and their combinations. The feature was extracted using MFCC, LPC and other analytical methods. In this paper, 35 audio features were extracted from the harmonium signal, which was then trained and tested by ANN with different combinations of features, which gives better accuracy. Some combinations like MFCC with CQT give 100% accuracy, while most others give above 90% accuracy. This method achieves the promising results on the self-generated database. The audio recordings were collected from professional harmonium artist in a varied environment.