An enhanced TEA algorithm for modal analysis

Turbo estimation algorithms (TEAs) for non-random parameters are able to yield high accuracy estimates by means of an iterative process. At each iteration, a noise reduction is performed by means of an external denoising system (EDS), which exploits the estimation results obtained at the previous step; the enhanced data are then input to the master estimation algorithm (EA) for next iteration. A basic TEA scheme has been previously proposed in the context of modal analysis, which makes use of the Tufts and Kumaresan (1982) algorithm as the master EA, and of a multiband IIR filter as the EDS. In this paper, two improvements of this basic scheme are proposed; the former implies a different design of the EDS, able to achieve better estimation accuracy while reducing the outlier probability; the latter permits the autodetermination of the number of modes making up the signal