Multi Activity Recognition Based on Bodymodel-Derived Primitives
LoCA '09 Proceedings of the 4th International Symposium on Location and Context Awareness
MARS: a muscle activity recognition system using inertial sensors
Proceedings of the 11th international conference on Information Processing in Sensor Networks
[MARS] a real time motion capture and muscle fatigue monitoring tool
Proceedings of the 10th ACM Conference on Embedded Network Sensor Systems
MARS: a muscle activity recognition system enabling self-configuring musculoskeletal sensor networks
Proceedings of the 12th international conference on Information processing in sensor networks
| Hi-index | 0.00 |
We present an experiment that investigates the usefulness of muscle monitoring information from arm mounted force sensitive resistors (FSR) for activity recognition. The paper is motivated by previous work that has demonstrated the feasibility of using FSRs for muscle activity monitoring (on leg muscles) and presented some initial signals related to distinct arm activities. We systematically investigate the performance of an FSR system on 16 different manipulative gestures and 2 subjects. The aim is to test the limits of the system, compare them to established sensing modalities (3D acceleration and gyro), and establish the value of combining FSR with other sensing modalities. We also present a hardware setup that addresses key problems that were identified in previous work: large variations in the attachment force and sensor placement accuracy issues. For all classifiers the overall accuracy of the FSR system is in the middle between the accelerometer (between 5% and 10% better) and the gyro (between 2% and 11% worse). Adding FSRs to another sensor improves the accuracy by 1% to 29%.