Bias compensation in the instantaneous frequency estimators based on the time-frequency representations and ICI algorithm

In this study, the author has proposed a technique for a bias compensation in the intersection-of-the-confidence-intervals (ICI) algorithm for the instantaneous frequency (IF) estimation. Algorithms from the ICI class are based on selection of the signal adaptive window length in the time-frequency representation-based IF estimators giving a trade-off between a bias and variance. The main difficulty in these estimators is sensitivity to a high noise influence that is producing outliers in estimates used for the ICI algorithm initialisation. Several ICI algorithm modifications are proposed in order to achieve robustness to the high noise influence. However, these techniques can suffer from the emphatic bias. The proposed bias compensation algorithm is based on the estimation of high-order IF derivative. It gives improvement in the mean squared error for more than 5dB with respect to the state-of-the-art variants of the ICI algorithm.