Phonetic characteristics of spontaneous speech in a total laryngectomized Italian speaker: Perspectives for speech enhancement algorithms

This paper describes the main phonetic features of an Italian L1 74 y. o. speaker (ESO01) after he endured total laryngectomy in 2015 with the complete removal of vocal folds due to five tumour masses. We offer an acoustic analysis of the spontaneous speech of this target speaker, in order to lay ground to the development of spontaneous speech enhancement and reconstruction algorithms for non-invasive aids. A semi-automatic analysis extracts formants’ values (F0, F1, F2, F3) on the midpoint and on 7 time-points, together with other acoustic cues. Our results show that our target speaker presents a low and rough voice, but his vowels are clearly differentiated. Furthermore, we find vocoid and air release to be extremely consistent in his acoustic characteristics during oesophageal phonation.