A new iterative algorithm for minimum-phase signal reconstruction from its autocorrelation

An iterative algorithm for minimum (maximum) phase signal reconstruction from its autocorrelation is presented. The algorithm is an extension of previous work by Quatieri and Oppenheim (1981). In particular, a good convergence speed is obtained even with no prior information on the first sample value of the signal to be reconstructed. A double time-domain constraint is imposed, and the knowledge of the autocorrelation function in both the time and the frequency domain is exploited. The algorithm shows a good convergence. However, it has two drawbacks: phase unwrapping must be performed at each iteration, and care must be taken in choosing the startup sequence for the iteration initialization.<
>