UpperMyo is an open six-muscle surface electromyography (sEMG) dataset for upper-limb action detection, temporal prediction, and cross-dataset transfer, evaluated through internal validation and paired external experiments. Analyses use 264 recording blocks from 24 healthy male participants, sampled at 1 kHz with 12-bit resolution; the nine-class dataset contains 2,640 four-second samples spanning eight actions and background. Internal evaluations cover recognition, channel ablation, and elbow-flexion/rest temporal classification. A convolutional neural network (CNN) achieved a macro-F1 of 0.9860 under nested leave-one-subject-out validation. Temporal classification yielded CNN accuracies of 89.5% and 75.3% at 0 and 2,000 ms backward offsets from an sEMG onset reference. Four external experiments compare transfer with target-only learning at identical labeling budgets for matched actions, new actions, activities without class mapping, and continuous grip-force estimation. WristMotion balanced accuracy increased by 14.21 percentage points with four labeled trials; mean flexor carpi ulnaris (FCU) grip-force R2 increased from 0.6431 to 0.6899 at the full budget. Transfer yielded gains across multiple tasks and budgets, with benefits varying by task and labeling budget. UpperMyo provides traceable data and evaluation evidence for sEMG action analysis and cross-dataset learning.
Wearable force-sensitive insoles enable quantitative gait analysis as a screening aid for neurodegenerative diseases (NDDs), yet prevailing machine learning pipelines give point predictions with no per-case reliability estimate, no intrinsic explanation, and no way to curate their own knowledge base. Case-Based Reasoning (CBR) mirrors clinical reasoning, but deployed healthcare CBR systems typically implement only partial R4 cycles, omitting Revise and Retain. We propose CogCBR, a sensor-driven framework that operationalizes the complete R4 cycle—Retrieve, Reuse, Revise, Retain—for gait-based NDD screening within Richter’s four knowledge containers, pairing weighted case retrieval with confidence-based clinical triage and a label-verified case-base maintenance policy. On the PhysioNet GaitPDB cohort, CogCBR attains an AUC of 0.861—statistically on par with the strongest tuned baseline under matched tuning, yet the only method evaluated that also provides confidence-based triage, case-based explanation, and longitudinal case-base maintenance, the last validated in a deployment-style streaming simulation. An independent-cohort evaluation on GaitNDD yields an AUC of 0.902; under a stricter cross-modality transfer, however, CogCBR does not exceed the strongest classical baseline, which is also reported. With sub-millisecond inference and a compact footprint, CogCBR suits resource-constrained wearable and edge-health platforms. Prospective longitudinal clinical evaluation and validation in pre-clinical cohorts are left as future work.