Subspace-based estimation of symbolic periodicities

In this work, we propose a novel subspace-based estimator of periodicities in symbolic sequences. The estimator exploits the harmonic structure naturally occurring in symbolic sequences and iteratively forms the estimate of the periodicities using a MUSIC-like formulation. The estimator allows for alphabets of different sizes, but is here illustrated using both simulated and real DNA measurements, showing a notable performance gain as compared to other common estimators.