Parallel frequency tracking with built-in performance evaluation

The problem of estimation of instantaneous frequency of a nonstationary complex sinusoid (cisoid) buried in wideband noise is considered. The proposed approach employs a bank of adaptive notch filters, extended with a nontrivial performance assessment mechanism which automatically chooses the best performing filter in the bank. Additionally, a computationally attractive method of implementing the bank is proposed. The new structure allows one to improve tracking results considerably, especially in nonstationary conditions. In terms of accuracy of frequency estimates, the proposed scheme outperforms existing ones considerably.