2. Corpus

2.1 The need for a corpus

In order to design new markers of sleepiness, we are proceeding in two stages:

1) Using previous systems proposed in the scientific literature, but also in collaboration with speech therapists, we design markers reflecting our intuition of the impact of sleepiness on the voice;

2) using statistical and/or machine learning methods, we select and validate these markers as biomarkers of sleepiness.

These two steps require the prior recording of the voices of drowsy and non-drowsy people, grouped together in a database – also known as a "corpus".

The following sections describe the development of the TILE corpus