Sleep staging by hyperdimensional dense networks

Leonardo Hernandez Cano, Alejandro Hernández-Cano, Ana Leonor Rivera

Abstract:

raditionally, sleep staging is done by medical experts, but computer aid will improve sleeping evaluation. We propose
a mathematically-motivated algorithm based on Dense Convolutional Networks that encodes polysomnography (PSG) recordings
into a very high-dimensional vector space to perform sleep-stage scoring. We emphasize the flexibility of our model as it provides a
framework to analyze single or multi-channel signals without relying on any statistical information about the dataset. To prove the
feasibility of our model we show results attaining comparable or better accuracy than current state-of-the-art models at a fraction
of the time and very limited training data.