Continuous Lower-Limb Joint-Angle Estimation from sEMG and IMU: The Conditional Value of Cross-Activity and Cross-Subject Data under Update-Matched Controls
Organizations: Conamur · Conamur, Palo Alto, CA, USA. · Institute of Neuroscience, Chinese Academy of Sciences · Institute of Neuroscience, Chinese Academy of Sciences, Shanghai, China. ·
Abstract
Intent-driven exoskeleton control needs a continuous, causal estimate of the wearer's joint kinematics and an answer to a second question: whose data should train the estimator for a new user? Borrowing the user's recordings of other activities, or other users' recordings of the same activity, is usually evaluated as "more data" versus "less data," which confounds the information a source carries with the optimizer updates it adds and hides negative transfer behind averages. We separate these effects for hip, knee, and ankle angle estimation from high-density sEMG and an IMU across six locomotion activities on all ten participants of a public dataset. A compact residual causal TCN-GRU decoder is frozen, and an update-matched replay control prices every added source against the same number of updates on the user's own data. A user's own other activities raise from 0.722 to 0.788, of which +0.048 is information beyond replay; stair descent is a reliably harmful donor and sit-stand-walk a reliably beneficial one. Other users' same-activity data help while the user has only the target activity () and half as much once own other-activity data exist (); at full calibration, donor value collapses after one to two target records. Per-target gains from the two sources correlate, and leave-one-participant-out gates recover most of the cross-subject value only at low calibration. Data value is hierarchical, state-dependent, and directional; the bottleneck is deciding, without test labels, when to transfer.
Figures & tables
| State | RMSE ( ∘ ) | MAE ( ∘ ) | CCC | ||
| Target-only | 0.7222 | 6.681 | 5.146 | 0.896 | 0.853 |
| Replay (update-matched) | 0.7400 | 6.368 | 4.910 | 0.903 | 0.858 |
| Pooled (+ own other activities) | 0.7880 | 5.631 | 4.399 | 0.929 | 0.893 |
| State | Phase-1 source | RMSE ( ∘ ) | MAE ( ∘ ) | ||
| T | replay | 0.7444 | 6.323 | 4.915 | 0.905 |
| T | cross-subject | 0.7969 | 5.452 | 4.287 | 0.931 |
| W | replay | 0.7823 | 5.605 | 4.394 | 0.926 |
| W | cross-subject | 0.8100 | 5.220 | 4.096 | 0.933 |
| no target data | cross-subject only | 0.2098 | 10.764 | 8.797 | 0.805 |
| Regime | Policy | Selected | [95% CI] | ( ∘ ) [95% CI] | Pos. / neg. a | Worst participant |
| Own other activities (Stage 2) | Always pool the five other activities | 60/60 | – / 24 | |||
| Fixed single donor (chosen on training participants) | 60/60 | – / 23 | ||||
| LOPO compatibility safe gate | 20/60 | 11 / 9 | ||||
| Oracle: best positive single donor b | 53/60 | 53 / 0 | ||||
| Other users, state T (Stage 3) | Always cross-subject pretraining | 60/60 | 40 / 20 | |||
| LOPO safe gate | 28/60 | 20 / 8 |
Transfer Learning
Transfer learning studies how models reuse knowledge across tasks, domains, or modalities when target data or computation are limited.
See all 449 papers