A preliminary study on the accuracy of wireless sensor fusion for biomotion capture

Abstract

A 2.4GHz low-power wireless platform based on the Texas Instruments™ CC2431 System-On-Chip (SOC) transceiver is used to conduct a preliminary investigation of the accuracy of a 3-axis accelerometer fused with a linear encoder for capturing high-speed motion. Calibration of the fused sensors is accomplished by extracting the gravity related term from the accelerometer when it is under static conditions. When moving at acceleration greater than 1.1g, the wireless platform switches to the linear encoder for local joint angle data. To ascertain the concept of sensor calibration when the sensor is under static conditions, experiments that compare results from a high-speed motion analysis system (Silicon Coach™) and the combination of the accelerometer and linear encoder is carried out. Experimental results for pronation and supination of the forearm that has acceleration less than 1.1g show that the accelerometer and linear encoder produces a low root mean square (RMS) error of 5.3° and 6.6° respectively. Correlation coefficient for the accelerometer and linear encoder for the above experiment are 98.7% and 98.9% respectively. A similar experiment where the maximum acceleration exceeds 1.1g produces a RMS error of 10.5° and 6.9° for the accelerometer and linear encoder respectively. Correlation coefficients for both sensors are approximately 98.2%. This study shows that linear encoder as a joint angle measurement device is preferred for both high and low acceleration motion while the errors associated with extracting gravity related tilt from the accelerometer increases with increasing acceleration. ©2008 IEEE.