Abstract:
Interest in examination of senescence voice and its phonetic traits are at the cornerstone of this thesis work. While age-induced changes had been shown to have a detrimental affect on the phonetic system, the analysis methods were most often invasive in nature (i.e. Electroglottography). An alternative approach involves the derivation of the glottal flow waveform, which is complementary to the behaviour of the vocal folds during cyclic vibration. In the past, the advancement of glottal analysis was hindered due to its computational complexity. Over the past two decades, however, advancements in speech processing operations allow for re-liable and efficient extraction of the glottal source. In turn, quantitative analysis of descriptive features of the glottal source provide a looking glass into physiological changes with advanced age. The contribution of this thesis work is two-fold. First, the design and implementation procedures for a voice analysis toolbox in R is provided. The R environment was chosen as it links to the emuR framework, enabling efficient analysis of large volumes of speech data. This toolbox implements popular speech processing algorithms in a computationally efficient manner for the extraction and parametrization of the glottal source. Using the popular iterative adaptive inverse filtering algorithm, six time- and amplitude-domain parameters were investigated, including the newly proposed open quotient criterion, OQsub50. Second, the impact of age, vowel, and age-vowel interactions are dis-cussed. The findings show correlation between senescence voice and the shape of the glottal pulse. Furthermore, vowel types were found to alter the glottal signal, implying coupling of the glottal source and the vocal tract filter. Finally, ageing had an affect on the production of vowels, implying results cannot be generalized across different vowels. This thesis work propels glottal signal analysis to new heights, highlighting the robustness of the extraction process, its correlation to speech production, and its usefulness for voice analysis.
