Implementation and Evaluation of a Sensor-Based Hand Grip Trainer

This thesis aims to replicate and critically evaluate a previously conducted project on a sensor-based hand grip trainer to assess the reliability of its methods, identify potential improvements, and compare outcomes. The work involved assembling a signal amplification circuit, constructing a 3D-printed prototype, and using a microcontroller to acquire and transmit sensor data. Mathematical tools such as the Least Squares algorithm and Kalman filtering were applied to estimate parameters and the state of a simplified muscle model, with particular focus on analyzing signal noise. Adding a voltage divider reduced signal noise and improved the accuracy and consistency of Least Squares estimates, especially when combined with a moving average filter. The Kalman filter, however, showed similar limitations as in the reference project, likely due to a simplified implementation and challenges in estimating total muscle mass. These findings are discussed alongside the assumptions and limitations of the physical and mathematical models, and potential improvements are suggested to support further development. Overall, the work contributes to enhancing the reliability of sensor-based grip assessment, with relevance for health monitoring and rehabilitation applications.