Research Program
Drowning in swimming pools refers to water entering the lungs in an artificial, enclosed swimming facility, causing hypoxia, suffocation, and other injuries. With the increasing popularity of indoor swimming, drowning has become a significant cause of unintentional death worldwide in recent years, with children having the highest drowning mortality rate in pools.
First edition
According to research, although swimming pools are equipped with lifeguards, the high number of swimmers often makes it impossible for lifeguards to monitor the status of all individuals at all times. To address the risks posed by this issue, drowning detection devices have been proposed. However, most of the existing drowning detection devices on the market rely on visual detection combined with intelligent analysis. The current image detection technologies, however, struggle to effectively handle issues such as obstruction caused by the large number of people in the pool, as well as problems like waves, refraction, and foggy mirrors underwater, which hinder proper analysis. Thus, I tried to use accelerometer to detect swimming strokes. This is my first edition of Arduino watch with simple code to test.
Second edition
I first implemented the Fast Fourier Transform (FFT) algorithm through programming to extract frequency-domain features from acceleration signals. I then combined it with the Support Vector Machine (SVM) algorithm to construct a swimming stroke classification model to detect abnormal motions of the swimmer. After finishing the coding I wore the second edition of the watch to collect testing datas.
Third edition
Because using SVM could only distinguish between normal and abnormal swimming postures, I switched to using the LSTM (Long Short-Term Memory) time-series neural network model. Additionally, I hand-soldered the circuit board, learned 3D modeling, and 3D-printed the third edition watch case, ensuring its airtightness.
