Visualisation of perturbed vowel phonation using dynamic fundamental harmonic normalisation

In larynx cancer patients the monitoring of vocal fold behaviour can be achieved using glottal waveforms for vowels captured as trans-larynx impedance time series via an electro-laryngograph. The impedance signals are sustained for up to 4 seconds, digitised at 20 kHz, and divided into 1000 point frames. An estimate of the instantaneous power spectral density PSD is computed from of each frame. However, detailed spectral differences between normal and aberrant vowel phonation are obscured by dynamic variations in fundamental frequency f/sub 0/. To separate fundamental and harmonic behaviour, f/sub 0/ is extracted from a frame and used to normalise spectral powers and to transform the entire frequency range into floating point multiples of the fundamental frequency. FHN dynamic PSD can then be presented as shaded surfaces to the clinician for evaluation. Examples are given for normal volunteers and a patient assessed by speech and language therapists pre-treatment, then at 6 and 12 months post treatment.