Projection Filter Based Framework Towards Improvement of Similarity Index Between Notes of Synthesizer and Indian Classical Instrument (Flute)

Abstract

In this paper we propose an approach of signal synthesis which modifies the input signal to attain the quality of reference signal. The paper proposes the hybrid approach of adaptively modifying the input signal to attain the similarity of reference signal. Different approaches of signal modification using cascading configurations of adaptive filters prominently incorporating Affine Projection filters are presented with formulation the most optimum approach is presented. The adaptive filter approaches are demonstrated with inclusion of Empirical Mode decomposition technique to arrive at the most effective framework. The quality of the synthesized signal is evaluated from different experts to confirm the effectiveness of the proposed approach. The proposed framework is tested with the musical domain where the reference signal is recorded from flute and the input signal is recorded with electronic synthesizer. The effectiveness of the framework is also examined by calculating different mathematical distances and regression plots. The framework proposed to be a generic framework for signal modification as it is also tested with the signals recorded from sitar. To the best of authors' knowledge no work has been reported towards incorporating qualitative naturalness in the synthesized signal. This paper proposes solution for the same.