<![CDATA[Low-cost inertial measurement unit (IMU) sensors can be used as footpods for virtual running and sensor learning, but speed estimation is sensitive to the sensor, mounting position, and user gait. This study develops a self-calibrated IMU footpod for estimating speed and cadence from foot motion. An ESP32-based prototype with a 6-axis IMU was mounted on the instep. The data were processed using quality control, 20-second windowing, gravity compensation, stance detection, zero-velocity update, feature extraction, and regression calibration. Eight recording sessions produced 40 valid windows at approximately 97 Hz, with 0% packet loss and no sensor saturation. Raw ZUPT estimation yielded an MAE of 3.242 km/h, whereas in-sample calibration reduced the MAE to 1.022 km/h. Cross-subject and cross-device transfer errors support the need for personal calibration. The pipeline also provides a practical learning medium for IMU calibration, filtering, drift, and wearable systems.]]>
Copyrights © 2026
