Computationally efficient IAA-based estimation of the fundamental frequency

Optimal linearly constrained minimum variance (LCMV) filtering methods have recently been applied to fundamental frequency estimation. Like many other fundamental frequency estimators, these methods are constructed using an estimate of the inverse data covariance matrix. The required matrix inverse is typically formed using the sample covariance matrix via data partitioning, although this is well-known to adversely affect the spectral resolution. In this paper, we propose a fast implementation of a novel optimal filtering method that utilizes the LCMV principle in conjunction with the iterative adaptive approach (IAA). The IAA formulation enables an accurate covariance matrix estimate from a single snapshot, i.e., without data partitioning, but the improvement comes at a notable computational cost. Exploiting the estimator's inherently low displacement rank of the necessary products of Toeplitz-like matrices, we form a computationally efficient implementation, reducing the required computational complexity with several orders of magnitude. The experimental results show that the performance of the proposed method is comparable or better than that of other competing methods in terms of spectral resolution.