Image Sequences as Necessary Supplement to a Pathological Voice Data Base

Voice corpora of pathological voices are generally based on high-quality audio recordings. These are supplemented with additional information describing the general recording conditions and the individual subjects. While an acoustic analysis cannot be expected to specify certain medical diagnoses it may reflect the glottal phonation conditions. Two examples shall illustrate this point. First, voice quality generally improves for patients undergoing logopedic therapy. These changes are reflected in the acoustic analysis by means of the hoarseness diagram while the diagnosis remains unchanged. Second, a statistically significant differentiation of the acoustic measures is achieved for subjects with laryngeal paralysis only if the phonation mechanism are taken into account. These examples demonstrate the necessity to add visual recordings to an audio data base of pathological voices in order to allow a flexible and comprehensive use of the voice corpus.