Fitness motion recognition with Android wear
summary
An Android Wear app uses the device accelerometer magnitude and an online peak detection algorithm (PeakDet) to count repetitions as they happen, reliably detecting 10 push-ups from real watch data. Key parameters: Delta ≈ 40 controls sensitivity and needs per-user calibration, sampling set via SENSOR_DELAY (FASTEST/GAME/UI) affects accuracy and battery, gyroscope is avoided for power, and processing runs on-watch with optional phone offload.
AI-written summary · about
chapters (9)
- 0:00Intro
- 2:00Exercise recognition
- 4:50Sensors
- 6:10Accelerometer
- 7:45Magnitude
- 10:00Peak Delta
- 13:20Battery Efficiency
- 15:20Algorithm Selection
- 16:20Sample app
related talks
23mCounting exercise repetition using live camera video
06mTesting Wear OS fitness apps without breaking a sweat
45mAndroid Extreme Wearables
27mDissecting ActivityRecognition with Multi-Sensor Simulation and Its Practical Deployment
14mCreating helpful fitness experiences with Health Services and Health Connect- 06mTesting Wear OS fitness apps without breaking a sweat
09mAn Activity Recognition System for Taking Medicine Using In-The-Wild Data to Promote Medication ...- 14mCreating helpful fitness experiences with Health Services and Health Connect