Adaptation convergence of spectral estimation based on Pisarenko harmonic retrieval

Proposed by Thompson in [1] was a technique for adaptive spectral estimation, particularly suited for tracking narrow-band signals in broad-band noise. This technique represented a data-adaptive variation of the block-oriented Pisarenko analysis technique. The algorithm departs from similar adaptive approaches based on linear prediction in that its estimate of line frequency is unbiased with the level of additive white noise. The original work offered a qualitative description of behavior and did not attempt the analysis of convergence characteristics necessary for practical application. This paper presents a preliminary study of transient and steady-state behavior, indicating conditions for successful application. While the characteristics parallel those of adaptive linear prediction in a general sense, certain significant variations are evident in both global and local convergence behavior.