Detection of Urination Using Machine Learning and Acoustics

Various factors, such as hydration levels, urinary tract diseases, prostatic hyperplasia, neurological disorders, medications, diabetes, and renal failure, can affect urination. This article explores the possibility of continuously evaluating urinary health using IoT technology by employing a contact microphone attached to the outside of the toilet bowl to record the acoustic patterns of urination for subsequent analysis. The performance of several algorithms for detecting urination patterns was investigated. Acoustic recordings were divided into segments of different sizes, from which 11 features were extracted. Support Vector Machines (SVM) were then used to assess the algorithm's effectiveness with various combinations of features and segment sizes. The aim of this study is to investigate the effectiveness of different methods for detecting acoustic patterns of urination, providing a range of algorithmic alternatives adaptable to the available processing capacity for detection.