Machine Learning Model Based Digital Hardware System Design for Detection of Sleep Apnea Among Neonatal Infants

Abstract

This paper presents a fully integrated machine learning (ML) based hardware system for detection of sleep apnea among infants in neonatal intensive care unit (NICU). The system is comprised of a PVDF sensor and a pulse oximeter to acquire breathing signal and oxygen saturation level, respectively, representing the input data. Accuracy rate of this system is over 85 percent with low error loss. The trained ML model has been developed in digital hardware design platform by translating each component into corresponding logic block. Estimated power consumption budget of this system is below 9W. This model can be adopted in future low-cost ML on-chip biomedical system design for apnea detection.