The pitch determination algorithm presented in this paper derives pitch periods from speech signals in time domain. It applies a crude approximation of an inverse filter which is realized as a two-step nonrecursive digital filter. In the first step, a low-pass filter cuts off the higher formants; the second step compensates the formant F1 by a tie-variant zero. The output signal of the filter shows a time structure similar to the glottal waveform: one principal peak per pitch period which is detected and marked as preliminary pitch period limitation. These markers, finally, are shifted into an optimum phase position. The algorithm requires the presence of the first harmonic in the signal; in bandlimited signals, this harmonic is reconstructed by nonlinear signal weighting between the two filter steps.
(See document for exact affiliation information.)
Click to purchase paper as a non-member or login as an AES member. If your company or school subscribes to the E-Library then switch to the institutional version. If you are not an AES member Join the AES. If you need to check your member status, login to the Member Portal.
