Abstract
A robust torque estimation system is developed for soft sensor suits embedded with Electromyogram (EMG) and a variety of sensors for measuring human muscle activities. The sensor suits serve as a man-machine interface for human power amplification. This paper focuses on the EMG part of the sensor suits to study the fundamental issue of torque estimation based on the sensor readings. A robust torque estimation system is developed in this study based on neural networks, where an universal database, rather than individual database for each operator, is proposed and as a result, the calibration time is dramatically decreased. The proposed system can also compensate for sensor positioning errors and individual differences.
