Deriving Effective Human Activity Recognition Systems through Objective Task Complexity Assessment
summary
Framework maps each HAR task to a vector of objective complexity measures—feature overlap, linear separability and mixture identifiability—normalized to (0,1) and visualized as radar plots. PCA reduces redundancies across 23 HAR datasets, producing a similarity map that yields six clusters (specific target behaviors, off‑body, arm gestures, repetitive, simulated, device‑embedded). A leave‑one‑task‑out validation gives three actionable outcomes: off‑the‑shelf works, heavy engineering required, or pipeline must be built from scratch; the method also provides pre‑deployment recognition estimates.
AI-written summary · about
related talks
06mA Systematic Study of Unsupervised Domain Adaptation for Robust Human-Activity Recognition
11mDetection and Differentiation of Obstacles in Repeated Adaptive Human-Computer Interactions from ...
06mRobust Unsupervised Factory Activity Recognition with Body-worn Accelerometer Using Temporal ...
28mDeep Learning-Based Human Activity Recognition for Physical Skills...
48mInterpretability - now what?
06mWeakly Supervised Multi-Task Representation Learning for Human Activity Analysis Using Wearables
05mMETIER: A Deep Multi-Task Learning Based Activity and User Recognition Model Using Wearable Sensors
41mAI applications in autonomous driving