Speaker: Matt Morrison
Research Advisor: Dr. Musa Audu
In person in Wickenden 321
Title: Characterization & Reorientation of Accelerometers for Enhancing Trunk Stability After Spinal Cord Injury
Abstract: There are currently an estimated 300,000 people in the United States living with spinal cord injury (SCI), with about 18,000 new cases of injury occurring each year. A vast majority of injuries result in paralysis that cause complete or partial loss of sensory and motor function below the level of injury. Trunk stability is consistently rated as one of the top priorities for recovery in this population. Control of implanted neuromuscular stimulation using trunk tilt measurements as a feedback signal has been shown to greatly improve seated trunk stability. A limitation of current control systems, however, is that these systems are based on tilt measurements from a single, externally placed sensor, making it difficult for constant use outside of a laboratory setting. To counteract these limitations, our team has begun work with the Networked Neuroprosthesis (NNP), which contains a collection of implanted modules capable of outputting acceleration measurements in 3 dimensions. A limitation of the NNP system is that the orientations of the implanted sensors are unknown. These unknown orientations make it difficult to obtain relevant orientation information. In this study, I investigate the Gram-Schmidt (GS) method for orthonormalization by using model data to verify the algorithm’s robustness. In the model, the GS algorithm is applied to clean signals as well as signals with a variety of corruptions. In addition, after confirming the utility of the GS method in-silico, I use this method to reorient signals from implanted NNP accelerometers to align them with the principal axes of the trunks of two subjects whose tilt has been measured in recent experiments. These processed outputs are then compared to outputs from a Vicon motion capture system, which acts as a gold standard for measuring tilt. The results of both the simulated and experimental studies show the effectiveness of the GS method in providing accurately reoriented measurements of tilt for future experiments. I also discuss preliminary results regarding the use of sensor fusion for subjects with multiple implanted sensors as a method for minimizing the variation between accelerometer and motion capture signals.