Smart Living Products
ISDN2001/2002: Second Year Design Project
Independent Study
IMU Motion Tracking & Stability Analysis
By Alan - LI, Shirui
Overview
Developed an IMU-based motion tracking and stability evaluation system for Table Archery using the ICM-42688 6-axis sensor integrated with ESP32S3 and Unity. The study focused on improving real-time camera control and creating a quantitative stability scoring mechanism by reducing drift and jitter during gameplay. Multiple signal-processing techniques were implemented, including gyroscope bias calibration, deadzone filtering, spike suppression, and signal smoothing.
System Workflow
ICM-42688 acquires 3-axis acceleration and gyroscope data
ESP32S3 performs real-time preprocessing and filtering
Motion data transmitted through WiFi/UDP to Unity
Unity converts processed signals into first-person camera movement and stability evaluation
Results
Observed Challenges
Long-term gyroscope drift accumulated over time
Small sensor fluctuations introduced camera jitter
Sudden spikes occasionally caused unstable movement
Implemented Improvements
-
Gyroscope bias calibration reduced baseline offset
-
Deadzone filtering removed minor unwanted movement
-
Extreme value clamping reduced sudden camera jumps
-
Signal smoothing improved motion consistency and responsiveness
Stability Scoring Mechanism
To quantify aiming performance, a stability scoring system was developed using the magnitude of gyroscope motion:
Total motion intensity calculated from 3-axis gyroscope data
Average movement magnitude measured during aiming duration
Lower motion values corresponded to higher stability scores
Stable aiming behavior contributed directly to gameplay performance and final scores
Outcome
-
IMU drifting remained visible during extended operation and could not be completely eliminated
-
Drift accumulation was constrained and reduced compared with raw sensor output
-
Camera movement became smoother and more controllable
-
Stability scoring successfully differentiated stable and unstable aiming behavior
-
Overall interaction quality and gameplay feedback became more reliable
Key Insight
Raw sensor data naturally contains noise and drift. Rather than completely eliminating these effects, the study demonstrated that combining filtering techniques with motion-based scoring can transform imperfect sensor signals into practical and meaningful user interaction.