ABSTRACT
In this research, we provide a detailed analysis of five adaptive
algorithms to attenuate hand tremors during writing. The evaluated
algorithms included Filtered Least Mean Squared (Fx-LMS),
Filtered Normalized LMS (Fx-NLMS), Hybrid Fx-LMS&NLMS, Recursive
Least Squares (RLS), and Kalman Filter. We have conducted
simulations to assess the performance of these algorithms using
the NewHandPD dataset, which contains hand tremor signals from
31 patients. Our results show that the mean squared error (MSE)
values of -38 dB for Fx-LMS, -42 dB for Fx-NLMS, -44 dB for Fx-LMS
and NLMS, -53 dB for RLS, and -50 dB for the Kalman Filter. RLS
had the lowest MSE and superior adaptation. On the other hand,
the Kalman Filter demonstrated faster convergence to the steady
state, which is six times faster than RLS.