Objective characterization of oesophageal voice supporting medical diagnosis, rehabilitation and monitoring.

Otolaryngologists use computational tools in the objective diagnosis of vocal folds pathologies by means of a set of acoustic parameters among others. This can be achieved in the cases of slight pathologies, but it does not exist any commercial software suitable for severe degradations of speech, as they are the oesophageal voice of laryngectomees. The present article shows a high-accuracy algorithm for the detection of the periodicity cycles of both oesophageal and laryngeal voices with low quality which allows the accurate and automatic estimation of pitch, jitter and shimmer. As the proposed algorithm works also with slighter pathologies, it is a useful contribution which allows doctors to perform an objective control during rehabilitation and monitoring stages. Thus, a patient can be controlled during oesophageal voice learning stage and it can also be saved a medical record with the results of the acoustic parameters' measurements in order to detect possible relapses.