Outlook and developments

From a technical point of view, future work on this task includes the development of new descriptors based on reading pauses (duration and location in the text), in order to take into account a third aspect of reading quality in drowsy patients.

Next, the development of a high-performance classifier (e.g. based on deep learning techniques) should enable this system to be deployed for clinical application.

From a methodological point of view, there are still many challenges to be met, such as switching to spontaneous speech, closer to real conditions of use, or moving from unidimensional detection to a multidimensional approach, integrating different aspects of sleepiness, or even several symptoms/pathologies simultaneously.