Detecting breathing rates and depth of breath using LPCs and Restricted Boltzmann Machines

This paper presents the use of a Restricted Boltzmann Machine to develop an unsupervised machine learning approach to process breathing sounds to predict breathing rates and depth or length of breaths. Breath detection and monitoring has been the subject of several studies involving the health monitoring of patients on respirators. We are proposing to extend the use of non-invasive techniques to provide measures of physical exhaustion or activity. The level of activity or exhaustion could be used to prevent accidents or manage exposure to physically demanding environments such as firefighting or working underwater. (C) 2018 Elsevier Ltd. All rights reserved.