Separation and analyzing of harmonics and inter-harmonics based on single channel independent component analysis

Abstract

The paper proposes a new method to separate and analyze harmonics and inter-harmonics on the basis of single channel blind source separation. Without any prior knowledge, a multichannel signal was acquired through a cycle-spinning method. After detecting and reducing dimension using principal component analysis, Independent Component Analysis is then applied to separate the independent harmonic and inter-harmonic sources. The independent parts are then clustered according to their frequency by using k-means, and mapped back to the observation domain. MATLAB simulations indicate that, as long as the sources are reasonably spectrally disjointed, the proposed method can effectively separate out the independent harmonics and inter-harmonics; thus, we can analyze them with fast Fourier transform-based method respectively. Simulation results show that the proposed algorithm has better estimation accuracy under high level of noise and is well suited for small-amplitude harmonic or inter-harmonic detection and analyzing.