Melodic Filtering for Indian Classical Instrumental Music

In this paper, various melodic noise filtering techniques viz. Adaptive Noise Cancellation, Spectral Methods and Deep Learning algorithms have been employed to filter music signals corrupted with additive Gaussian white noise. The noise reduction problem has been formulated as a filtering problem which is efficiently solved by using the LMS, NLMS and RLS methods in adaptive filtering and as a spectral problem solved using spectral subtraction and spectral gating techniques. Various performance criteria were considered to examine and analyse these methods. Results of the analysis shows that the LMS algorithm was the best based on all performance parameters. Adaptive filtering makes use of a desired signal as reference. However, without such reference signals, deep learning algorithms show better overall performance.