Parameter optimization for wheeze detection

• 主要作者: Yong, Clarissa Mei Hui
• 其他作者: Ser Wee
• 格式: Final Year Project
• 語言: English
• 出版: 2019
• 主題: Engineering::Electrical and electronic engineering
• 在線閱讀: http://hdl.handle.net/10356/78801
• 機構: Nanyang Technological University

Wheeze is an abnormal breath sound which occurs when airways in the lungs have been narrow and produced a continuous high-pitched whistling noise. Over the last few years, stethoscope has been used to analysis respiratory sound, but it requires interpretation time before diagnosis could be obtained. However, for cases such as several breathing problems which could lead to medical emergency and death diagnosis must be conducted within a short time span. Which resulted in more wheeze detection algorithm created and developed in the market. However, the acceptance of these algorithms highly depends on the diagnostic accuracy rate. Therefore, in order to improve the diagnostic accuracy results, different time sensitivity parameters have been studied in this study. In this study, the author would be focusing on Entropy Based Wheeze Detection (EBWD) algorithm to extract entropy features such as Entropy Difference and Entropy Ratio. Meanwhile, time sensitivity parameter optimization such as Length of Smoothing Filter, Ratio of Window Overlap, and Duration of Window would be analysed. Classification methods like K-Nearest Neighbour (KNN) and Support Vector Machine (SVM) were also studied. Through this study, an optimal linear classification result of 50% Window Overlap and 60ms Hanning Window achieved accuracy, specificity and sensitivity percentage of 93.8%, 100%, and 88.9% respectively.