Multitone tracking with coupled EKFs and high order learning

A multitone tracker is described using two basic principles in optimum frequency estimation: processing bandwidth depending on the distance from the estimate to the actual frequency values, and parallel estimates with inhibitory paths to ensure orthogonality between the enhanced tones. The first feature is provided by extended Kalman filters (EKFs), and the second one is achieved by a high-order rule for the learning of the inhibitory cells. It is shown that the independence between signals is linked to the high-order function of the learning process. The resulting multitone tracker seems to be a potential alternative to adaptive high-resolution methods or time-frequency tools