Multiple tone removal from short signal records using singular value decomposition

Multiple narrowband interferers in a wideband signal can be estimated using singular value decomposition (SVD) and then optimally subtracted to restore signal quality. This technique applies principally when the duration of the signal “snapshot” is small compared to the reciprocal of the instantaneous bandwidths of the interferers, so that the interference can be well approximated by a sum of tones over the collection interval. In this case, the SVD technique outperforms conventional techniques, such as conventional notch filtering or the constant modulus algorithm. The method relies on the fact that a sampled sum of tones, when rastered on some width, constitutes a matrix of low rank