Ultrasonic Raytracing Simulation Method for Data Augmentation to Surveil the Bathroom with Digital Twins

Bathroom accidents are common, requiring prevention and detection methods. Advances in computer-aided engineering and machine learning enable the creation of digital twins for analysis and productivity. A method using ultrasonic depth images and machine learning for human position estimation is proposed for personal security. A 16-ch microphone array captures reflected waves, generating distance-based images and converting them into depth and binary segmentation images. In this study, to advocate the lack of measurement data for input, we implement augmented data using the simulation method. Evaluation shows, our proposed method managed to improve the machine learning results with an average mean average error of 9.73 cm.