Answers are grounded in this paper and cite their sources.
Estimating the tilt angle of a body from a low-cost inertial sensor requires combining an accelerometer, which is noisy but free of drift, with a gyroscope, which is smooth but drifts. This paper compares gyroscope integration, accelerometer-only estimation, a complementary filter and a two-state Kalman filter (angle and gyroscope bias) in a simulation of a body oscillating by ±40° with a gyroscope bias of 0.5 °/s, accelerometer noise of 3° and a 2 s linear-acceleration disturbance. After tuning on separate runs, the Kalman filter gave an RMSE of 0.29° before the disturbance, against 0.39° for the complementary filter and 3.01° for the accelerometer alone. During the disturbance the Kalman filter was less affected (7.6° versus 12.2°), but afterwards it needed time to recover and had a larger error (1.78°) than the complementary filter (0.42°). The study is a one-axis simulation and the ranking depends on the chosen disturbance and tuning. This is a simulation study; all results are outputs of the deposited Python code with assumed parameters. Source code is attached.
No takes yet. Share an insight, caveat, or question.
Merdanow et al. (2026) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: