Isolating Road Wear from Cabin Noise
How our software isolates vehicle suspension movements from engine noise.
The raw accelerometer reads gravitational forces across three dimensions ($a_x$, $a_y$, $a_z$). To isolate true pavement anomalies, we first remove the baseline Earth gravity vector ($1.0g$) using a continuous low-pass gravity filter:
g_vector(t) = α · g_vector(t-1) + (1 - α) · a_raw(t)
a_isolated(t) = a_raw(t) - g_vector(t)Where $\alpha = 0.98$ acts as the low-pass smoothing factor. Pothole spikes are identified when the isolated vertical magnitude exceeds a dynamic threshold $T_{puka}$ calculated using running statistics:
T_puka = μ_accel + k · σ_accelSensor Cradle Rigid Mounts
Because smartphones vary in vehicle placement, rigid cradle installations are critical. Loose dashboard mounts introduce cabin rattling which mimics high-g pothole shocks.
- Mounting Preference: Solid glass suction cradle or bolted dashboard brackets.
- Excluded Setups: Phone placed loosely in cup holders, passenger seats, or coat pockets.
- Device Calibration: Run an initial 30-second flat drive calibration to determine the vehicle suspension variance vector.