An Iterative Algorithm for Signal Reconstruction from Bispectrum

This paper presents a procedure to reconstruct a (minimum or nonminimum phase) discrete-time signal from its bispectrum. The algorithm is iterative and based on using the method of projections onto convex sets (POCS) in an inner product space in which the vector space consists of real sequences, and vector addition is defined in terms of the convolution operation. Prior information, such as an energy bound on the signal, can be incorporated into this algorithm. An algorithm that employs energy information was found to produce better results in simulation than one that ignores energy constraints. © 1991 IEEE