Voice parameter extractor using oral airflow

Abstract

A technique for extracting voice parameters from oral or combined oral and nasal air flow involves having a test subject or patient pronounce a vowel sound that has a high first formant into a pneumotachographic device which produces an output voiced speech waveform that includes the mean air flow. This output signal is appropriately filtered to remove the formant frequencies and produce a smoothed glottal waveform approximation. From the pneumotachograph output waveform and smoothed glottal waveform are derived a peak value P, a minimum value L, glottal wave period TO and average value M for the glottal waveform. The amplitude and wave center value can readily be calculated, and from these the open quotient QO can be derived which represents the ratio of the duration of the open portion B of the glottal wave cycle to the entire period TO thereof. These values can be recorded directly, or can be used to synthesize a replica of the subject's glottal airflow for either one typical vocal fold vibratory cycle or a number of successive cycles. The technique can be carried out using digital techniques.