Subspace-based fundamental frequency estimation

In this paper, we present a subspace-based fundamental frequency estimator based on an extension of the MUSIC spectral estimator. A noise subspace is obtained from the eigenvalue decomposition of the estimated sample covariance matrix and fundamental frequency candidates are selected as the frequencies where the harmonic signal subspace is closest to being orthogonal to the noise subspace. The performance of the proposed method is evaluated and compared to that of the non-linear least-squares (NLS) estimator and the corresponding Cramér-Rao bound; it is concluded that the proposed method has good statistical performance at a lower computational cost than the statistically efficient NLS estimator.