Lattice-structured schur-type elimination of narrow-band background disturbances from acoustic signals of vehicles in motion

Abstract

The paper deals with the problem of detection of vehicles in motion, using emitted acoustic signals. Due to the presence of background disturbances, the registered real-life signals contain a mixture of a source signal (emitted by the vehicle), and a corrupting signal (narrow-band “background noise” of bandwidth similar to the original signal). A new method of the background noise elimination is proposed in this paper. The method is based on the lattice-structured Schur-type orthogonalization/parametrization algorithms in which the so-called backward estimation errors, computed for the incoming signal, constitute an orthonormal basis. The proposed solution employs the orthogonal joint-process estimation filter which decomposes the incoming signal using an orthonormal basis, derived exclusively for the background noise. The effectiveness of the method is discussed as well as the noise cancellation results, also using real-life signals, are presented.