Nonlinear autocorrelation function (FNLAF)—Real-time pitch extractor
C. Garland · The Journal of the Acoustical Society of America · 1976
This paper describes a digital imprementation of a pitch-extraction algorithm which is based on bit-pattern comparison. The NLAF is a variation on autocorrelation analysis where, instead of correlating the input speech waveform, the autocorrelation is computed either on the sign of the signal or on the sign of the first derivative of the signal. After a review of the method, the pitch extractor implementation is described. Then, experimental results are presented when analysis is first performed on glottal waveform, and then on speech waveform. The comparison of the results leads to the necessity of preprocessing the speech waveform in order to eliminate the adverse effects of the frequency components generated by resonances within the vocal tract. This preprocessing is performed by use of a very simple digital fileteter which considerably decreases the probability of an erroneous determination of the pitch period.