Adaptive ESPRIT algorithm based on the past subspace tracker

Summary form only given. Sinusoidal modeling is a powerful tool for audio signal processing, which represents the signal as a sum of sinusoids whose frequencies may vary over time. In most applications, the estimation and tracking of multiple frequencies is achieved by means of the Fourier analysis. Concurrently, we propose a new frequency estimation and tracking method, based on two algorithms derived from the concept of signal subspace. An application to audio is presented.