Intelligent Workplace Activity Monitoring and Detection Using Self-Powered Triboelectric/Piezoelectric Sensor Augmented Machine Learning

In this letter, we propose a method for monitoring workplace activity using a self-powered, flexible triboelectric/piezoelectric tactile sensor. The flexible sensor is applied to the forefinger of the participants while they perform typical workplace activities like keyboard typing, mouse scrolling, scribbling, and finger stretching. The electrical output signal corresponding to triboelectric and piezoelectric sensors for each activity is recorded by a data acquisition unit and analyzed using an efficient convolution neural network (CNN) algorithm to precisely distinguish between the different activities. The classification results demonstrate that the proposed strategy of incorporating dual transduction mechanisms results in an accuracy higher than
$98\%$
, and the precision and recall of
$98.13\%$
and
$98.14\%$
, respectively. Due to the simple and inexpensive design of the sensor and the remarkable accuracy of the proposed machine-learning architecture, the reported work has immense potential in a wide range of workplace monitoring applications.