Frequency Detection By Linear Prediction Using Total Least Squares Criterion

This work is concerned with the optimization of Forward-Backward Linear Prediction in Total Least Squares sense. This allowed us to propose a new frequency estimation method, named TLS-FBLP. It offers better performance and is less sensitive to prediction order changes than FBLP. Also, the proposed method is better than the Tufts and Kumaresan's Modified FBLP at low prediction orders, and similar at high orders. The best performances of both methods are quite close.282382