Deep Learning-Based Human Activity Recognition for Physical Skills...
summary
Claims an operational pipeline: IMU time series undergo statistical feature processing and clustering (kernel determinant analysis), then an LSTM classifies actions; camera frames go through pose estimation into a temporal graph convolutional network or LSTM. Camera systems lack reliable 3D rotation while IMUs can require up to 17 sensors. Demonstrated use cases include rehab rep counting, fall detection and assembly-line error spotting; a 12-action confusion matrix shows high diagonal accuracy for well-defined actions.
AI-written summary · about
chapters (8)
- 0:00Intro
- 0:51Why Does Human Activity Recognition Matter?
- 3:25Human Activities are Analogous to a Written Language
- 7:33How Does HAR Help in Health and Wellness?
- 13:58On-Body vs Off-Body Sensing: A Case for Sensor Fusion
- 19:39Pipeline for Inertial Sensor-Based Human Activity Recognition
- 23:28Pipeline for Vision-Based Human Activity Recognition
- 26:33Conclusion: The Next Frontiers in Human Activity Recognition
related talks
06mApproaching the Real-World: Supporting Activity Recognition Training with Virtual IMU Data
06mUnsupervised Human Activity Representation Learning with Multi-task Deep Clustering
22mSeeing, Sensing and Recognizing Laban Movement Qualities
27mAnalysis of sports scenes using computer vision models
46mML Use Cases at Think Big Analytics
06mRobust Unsupervised Factory Activity Recognition with Body-worn Accelerometer Using Temporal ...
14mBack-Hand-Pose: 3D Hand Pose Estimation for a Wrist-worn Camera via Dorsum Deformation Network
10mProactive Sensing for Improving Hand Pose Estimation