Data Analysis for Detecting a Temporary Breath Inability Episode

摘要

This research is focused on a real world problem: the identification of a specific type of apnea disorder. Recently, a technique for evaluating and diagnosing of a certain type of apnea has been published carried out. In a brief, this technique proposes that a subject is given with two belts, to be placed on the thorax and on the abdomen, respectively; each belt includes a 3D accelerometer. In a sleep laboratory, the subject is monitored while sleeping and the apnea episodes are manually discovered and registered. Besides, during the test, the data from the sensors is gathered and segmented. The hypothesis of this study is that the diagnose of the apnea episodes can be accomplished using the data from a single 3D acceleration sensor. If successful, this technique for the diagnose of the apnea might reduce the costs of the tests as well as allow evaluating challenging cases, as those related with children or with elder people. This study focus on the time series (TS) analysis to extract the most relevant patterns corresponding with the apneas episodes.Focusing on the analysis of the TS, this study will apply a well-known TS technique to extract the most relevant patterns. The main contributions of this study are i) to determine if a previous step for estimating the posture is needed, which is a very important decision in the design of the embedded solution, and ii) the evaluation of the hypothesis of diagnosing by means of a single 3D accelerometer.