The study investigates how the source of normalization statistics affects the performance of wrist electrodermal activity (EDA) affect‑recognition models. Using the SAFE‑EDA convolutional network pretrained on expert artifact annotations, the authors compare models trained with normalization derived only from training subjects versus from the held‑out subject’s full recording. They find that pretraining improves macro‑F1 when using training‑only statistics, but the benefit diminishes when using the held‑out subject’s data, and that artifact supervision outperforms self‑supervised pretraining. Across multiple configurations, pretrained models generally perform better, though the interaction with per‑user normalization varies by dataset.
arXiv:2607. 27565v1 Announce Type: new Abstract: Pattern-recognition control promises a myoelectric prosthesis that responds to many intended gestures rather than one or two, but the promise has stayed in the laboratory.
arXiv:2607. 27568v1 Announce Type: new Abstract: Recognition accuracy obtained during a recording session does not persist when a user puts on the electrodes again after the electrodes had previously been removed.
Pattern-recognition control promises a myoelectric prosthesis that responds to many intended gestures rather than one or two, but the promise has stayed in the laboratory. A recogniser trained on one person rarely transfers to the next, and useful performance usually demands a fresh round of labelled calibration from the end user.
The paper evaluates six onboarding strategies for federated wearable models on five datasets using a leakage‑controlled protocol that fixes source checkpoints and separates calibration from evaluation. Results show that while average accuracy is high, person‑level performance can drop significantly, with some methods causing negative transfer for certain users. The study highlights that mean accuracy alone is insufficient and provides an auditable benchmark and failure map for future development.
arXiv:2606. 01896v1 Announce Type: cross Abstract: Generated (or synthetic) image data is increasingly used to augment or replace real training datasets when target imagery is scarce, expensive, or biased.
arXiv:2608. 15861v1 Announce Type: new Abstract: Fine-grained wrist activity recognition can support applications such as procedural step guidance and context-aware assistance, yet acquiring labeled data for every new task, user, and activity granularity remains a bottleneck.
The paper proposes Dynamic Influence-Weighted Distillation (DIW) to improve single-IMU activity recognition by leveraging a frozen four-IMU teacher during training. DIW assigns sample-wise gates to logit and feature losses, outperforming both supervised learning and fixed-weight knowledge distillation on the WEAR dataset with a macro‑F1 of 0.638451. The method enhances a right‑arm IMU model without altering the deployed sensor setup or the student network architecture.
The paper investigates how to personalize a cross-user surface electromyography (sEMG) encoder when only a few calibration repetitions are available. Four methods—prototypical adaptation, linear probes, scaled fine‑tuning, and full fine‑tuning—were evaluated across 77 subjects on two databases. Full fine‑tuning consistently achieved the highest accuracy, but a gradient‑free prototypical rule captured 52–78 % of the benefit without per‑user weight copies, enabling quick donning‑time personalization.
The study compares temporal deep learning models—Bidirectional LSTM, Temporal Convolutional Network, and Transformer—for physiological emotion recognition using two multimodal wearable datasets, WESAD and EmoWear. Experiments evaluate wrist-only, chest-only, and multimodal sensor configurations with participant-independent leave-one-subject-out cross-validation, and also explore ensembles, sensor ablation, sampling frequency, and saliency analysis. Results show that the best architecture varies by dataset, multimodal sensing consistently outperforms single-site configurations, and a 4 Hz sampling rate offers a cost-effective operating point.
The paper introduces the Individual Conformal Coupling Monitor (ICCM), a lightweight pre‑inference tool that detects structural ambiguity—when physiological signals that appear plausible individually form a pattern poorly supported by a person’s non‑stress baseline—in wearable stress classifiers. On the WESAD dataset, a Random Forest achieves high mean accuracy but fails entirely for Subject 14 due to weakened cross‑signal coupling near stress onset. ICCM quantifies subject‑specific coupling divergence and can route data to classify, defer, or abstain, reducing false positives slightly and withholding some misclassified windows, though it does not fully correct the failure.
The study evaluates how well a surface‑EMG gesture recogniser trained on intact‑limb data transfers to transradial amputees. Zero‑shot transfer fails; the model needs a few labelled repetitions from the new user to outperform a per‑user classifier, achieving a macro‑F1 of 0.779 versus 0.589. Training on a larger pool of intact subjects, or combining intact and amputee data, yields the best cross‑population performance.
arXiv:2607.12176v2 Announce Type: replace Abstract: Ambivalence and hesitancy (A/H) undermine digital behaviour-change interventions, and recognizing them automatically from video is the goal of the...