Determining speech pitch by modified least-squares periodic approximation

The pitch of a voiced speech segment may be estimated by minimizing the squared error in approximating the seg ment by the repeated weighted average of successive assumed pitch periods, after a "maximum-likelihood" solution due to Slepian. A modification is proposed which corrects two faults of the original formulation. Computationally the method works in the time domain. Preliminary results indicate that the pitch estimate is immune to band-pass filtering which suppresses the fundamental frequency itself.