{"paper":{"title":"From Muscle Bursts to Motor Intent: Self-Supervised Token Modeling for Heterogeneous EMG","license":"http://creativecommons.org/licenses/by/4.0/","headline":"AEMG pre-trains EMG signals as a cross-device physiological language using a contraction tokenizer to improve generalization in motor intent decoding.","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Huilin Yao, Kaikai Wang, Lin Shu, Zhenghao Huang","submitted_at":"2026-05-05T07:50:00Z","abstract_excerpt":"Surface electromyography provides a practical way to infer human movement intention from wearable muscle recordings, but models trained under a single acquisition setting often lose reliability when the user, session, electrode layout, or gesture protocol changes. This paper proposes AEMG, a self-supervised learning approach designed to extract reusable neuromuscular representations from diverse EMG sources. Eight public gesture datasets are first transformed into a shared signal format to reduce discrepancies in channel configuration, sensor topology, and recording protocol. Instead of relyin"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"Experiments demonstrate that AEMG improves the zero-shot leave-one-subject-out (LOSO) accuracy by 5.79-9.25% compared to six state-of-the-art baselines, and achieves more than 90% few-shot adaptation performance with only 5% of target user data.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That EMG signals can be tokenized into a linguistic structure (words from contractions, sentences from temporal patterns) via NCT without losing critical discriminative information across heterogeneous devices and subjects.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"AEMG pre-trains EMG representations by treating neuromuscular signals as language via a novel tokenizer and cross-device vocabulary, yielding 5.79-9.25% zero-shot LOSO gains and over 90% few-shot performance with 5% target data.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"AEMG pre-trains EMG signals as a cross-device physiological language using a contraction tokenizer to improve generalization in motor intent decoding.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"143d9325966281221d468dee7e6a75fa8c14d2d31290d6d5d236f06f6ebfe7fe"},"source":{"id":"2605.03462","kind":"arxiv","version":2},"verdict":{"id":"db9201a9-9e15-429d-bfd8-6cbe8b550ebd","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-07T17:08:50.895608Z","strongest_claim":"Experiments demonstrate that AEMG improves the zero-shot leave-one-subject-out (LOSO) accuracy by 5.79-9.25% compared to six state-of-the-art baselines, and achieves more than 90% few-shot adaptation performance with only 5% of target user data.","one_line_summary":"AEMG pre-trains EMG representations by treating neuromuscular signals as language via a novel tokenizer and cross-device vocabulary, yielding 5.79-9.25% zero-shot LOSO gains and over 90% few-shot performance with 5% target data.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That EMG signals can be tokenized into a linguistic structure (words from contractions, sentences from temporal patterns) via NCT without losing critical discriminative information across heterogeneous devices and subjects.","pith_extraction_headline":"AEMG pre-trains EMG signals as a cross-device physiological language using a contraction tokenizer to improve generalization in motor intent decoding."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2605.03462/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"doi_compliance","ran_at":"2026-05-19T15:23:21.643880Z","status":"completed","version":"1.0.0","findings_count":0}],"snapshot_sha256":"c64002bcf71f7bd357d7b69b0baa6c29f11e0112affd4a35a85b37c680569630"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}