A comparison of interpolation algorithms for gramophone record sound reconstruction

Gramophone records were the main recording medium for half a century and regained widespread popularity over the past couple of years. Being an analogue storage medium, vinyl records are subject to distortions caused by scratches, dust, high temperatures and extensive playback. This paper discusses a number of well-known interpolation algorithms that can be used to reduce the noise by reconstructing the distorted gaps, which include the nearest neighbour, cosine, polynomial, spline, Hermite and Fourier interpolation. The reconstruction accuracy of these algorithms were determined by using a large dataset of songs from eight different music genres. It was found that cosine and polynomial interpolation have the best mean performance with regards to reconstruction accuracy and execution time. It was also found that certain genres can be easier reconstructed than others.