Fast Implementation of Harmonic Music for a Known and an Unknown Model Order

In this paper we present a fast implementation of the subspacebased MUltiple SIgnal Classification (MUSIC) estimation criterion for harmonic signals with a known and an unknown model order, respectively. The proposed implementation improves the MUSIC criterion in two ways: First, we introduce a novel and fast implementation for evaluating the cost function of the MUSIC estimator involving only one FFT for known model order, and we extend it to the case for unknown model order. Second, we introduce an algorithm which use a low complexity subspace tracker and a gradient based minimization method depending on the minimum value of the MUSIC cost function. The performance gain obtained by the proposed algorithms is significant and enables real-time implementation for a known model order and faster computation in some cases for an unknown model order.