Analysis and Quantification of Acoustic Artefacts in Tracheoesophageal Speech

After total laryngectomy, the placement of a tracheoesophageal (TE) puncture offers the possibility to gain a new voice. However, the produced TE speech is known to have a lower quality and intelligibility. The goal of this paper is to identify and quantify the acoustic artefacts in TE speech. The advantage of this study is two-fold. First, the proposed measures can be used by speech therapists in voice rehabilitation sessions to assess the voice of the patient, to follow up his/her evolution and to design tailored exercises. Secondly, these artefacts have to be quantified and taken into account in synthesis methods aiming at enhancing TE speech. Four categories of acoustic artefacts are identified in this work: a lower periodicity and regularity of the phonation, and the presence of high-frequency and gargling noises. Each artefact is studied and compared to normal laryngeal speech recorded either for speech synthesis purpose or by elderly people. Results quantify the importance of each of these artefacts, and show a large disparity between TE patients.