IF estimation in multi-sensor scenario with unknown sensor array geometry

Abstract

An instantaneous frequency (IF) estimator for multi-sensor recordings has been developed for a scenario where the array geometry is unknown at the receiver. The method involves two stages. In the first stage, the IFs, as well as the mixing matrix (steering matrix), are iteratively estimated through the application of the ridge tracking method in combination with joint time-frequency and spatial filtering. In the second stage, the IF estimates are further refined by converting a multi-component IF estimation problem into a mono-component. For the N source signal, the IFs of all the components are re-estimated by removing N-1 components from the mixture signal. The process of re-estimation is repeated till IFs do not change in two consecutive iterations or a specified number of iterations have been reached. Experimental results indicate that the proposed method outperforms the state of art algorithms.