An algorithm for synthesizing signals from partial time-frequency models using the cross Wigner distribution

A weighted-least-square error criterion that optimally synthesizes a signal when only part of the desired time-frequency model is available is formulated for time-frequency synthesis. The partial model occurs frequently in situations in which the ideal synthesis model has been truncated by a time-varying mask in order to exclude unwanted signal components or when knowledge of the true time or frequency support of the ideal model is unknown. A cross-Wigner-distribution (CWD) synthesis algorithm is developed from the weighted-least-square error criterion and compared with existing Wigner distribution (WD) synthesis algorithms. The algorithm is also shown to be useful when interference or cross terms hinder the use of the quadratic auto-Wigner-distribution (AWD) for signal synthesis