Pedometer Step Detection: Filtered Acceleration Trace, Threshold and Cadence (2D)
The raw and filtered acceleration trace with the adaptive threshold and detected steps marked, the filter gain curve and counts against the truth. The same shared model as the 3D wrist scene, with numerical checks.
A wrist accelerometer signal passes through a moving-average low-pass filter and a threshold peak detector with a refractory time. Change cadence, step strength, noise, filter window and threshold to see steps counted, missed or invented, with the filter gain curve and cadence estimates by autocorrelation and DFT. A Blender-built wrist and watch with a live 3D trace and numerical checks including Parseval.
The raw and filtered acceleration trace with the adaptive threshold and detected steps marked, the filter gain curve and counts against the truth. The same shared model as the 3D wrist scene, with numerical checks.
2D · HTML5 Canvas 2D · 60 FPS target · runs fully client-side, no install