Activity Identification using Inertial Measuring Unit Advanced Sensor Network

Abstract

The need to keep track of the activity a miner is engaged in, is germane to the productivity and safety of the worker. Apart from this, it also serves as a possible early warning system if any irregular movements are detected. Existing tracking methods suffer from high energy consumption and the inability to identify a person based on their gait as they only identify movements such as walking or sitting. To solve this challenge, this paper presents a low power inertial measuring unit (IMU) based system capable of identifying various activities, specifically those that a worker would engage in while working in a mine. The system is extended to perform a gait analysis to identify the person performing the activity as well. Three sensor nodes were designed and etched onto a printed circuit board (PCB). Housings for the nodes were designed and 3D-printed. Firmware for the sensor node microcontrollers was developed in C to incorporate I2C data sampling with XBee API mode packet construction. A back-end program was then developed in C# to handle all the incoming data with the use of 2 neural networks. The test of the proposed system revealed a high degree of activity identification accuracy while the results obtained for the gait analysis revealed that the system can distinguish between different users with reasonable accuracy. The energy consumption test also revealed a satisfactory performance.