{"@context":"https://neupai.io/schema/v0.2","@type":"StructuredNewsArticle","identity":{"article_id":"samsung_20260814_health-foundation-model-wearable-ai","canonical_url":"https://news.samsung.com/kr/%ec%82%bc%ec%84%b1%eb%a6%ac%ec%84%9c%ec%b9%98-%ed%97%ac%ec%8a%a4-%ed%8c%8c%ec%9a%b4%eb%8d%b0%ec%9d%b4%ec%85%98-%eb%aa%a8%eb%8d%b8-%ea%b8%b0%eb%b0%98-%ec%9b%a8%ec%96%b4%eb%9f%ac%eb%b8%94-ai-%ec%97%b0","ai_url":null,"publisher":{"name":"삼성전자 뉴스룸","domain":"news.samsung.com","type":"online"},"author":"삼성전자 뉴스룸","published_at":"2026-08-14T08:00:00.000Z","updated_at":null,"language":"en","article_type":"press_release_based","originality":"self_produced"},"content":{"headline":"Samsung Research Unveils Wearable AI Research Based on Health Foundation Models","summary":"Samsung Research America has unveiled two health foundation model AI research projects, 'xMAE' and 'HiMAE,' which learn from wearable biosignals. Both studies were accepted at the ICML and ICLR conferences respectively, with xMAE outperforming existing methods in 15 out of 19 evaluation tasks, and HiMAE demonstrating the potential for real-time on-device analysis.","topics":["AI","healthcare","wearables","Samsung Research","foundation models"],"geography":["KR","US"],"entities":[{"name":"Samsung Electronics","canonical_id":"corp:kr:samsung-electronics","type":"company","role_in_article":"primary_subject","metadata":{"ticker":"005930.KS","parent":null}},{"name":"Samsung Research America","canonical_id":"org:us:samsung-research-america","type":"organization","role_in_article":"source","metadata":{"parent":"corp:kr:samsung-electronics"}},{"name":"xMAE","canonical_id":"product:kr:xmae-model","type":"product","role_in_article":"mentioned","metadata":{}},{"name":"HiMAE","canonical_id":"product:kr:himae","type":"product","role_in_article":"mentioned","metadata":{}},{"name":"ICML","canonical_id":"org:us:international-conference-on-machine-learning","type":"organization","role_in_article":"mentioned","metadata":{}},{"name":"ICLR","canonical_id":"org:us:international-conference-on-learning-representations","type":"organization","role_in_article":"mentioned","metadata":{}}],"claims":[{"id":"c1","statement":"Digital health researchers at Samsung Research America are conducting research on health foundation model AI that learns from wearable biosignals","as_of":"2026-08","as_of_explicit":false,"as_of_raw":"August 2026","source_type":"company_disclosure","comparison":null,"type":"fact","figures":null,"expiry_hint":null,"insight":null},{"id":"c2","statement":"The xMAE and HiMAE studies were accepted at the ICML and ICLR conferences respectively, recognizing their excellence","as_of":"2026","as_of_explicit":false,"as_of_raw":"2026","source_type":"company_disclosure","comparison":null,"type":"fact","figures":null,"expiry_hint":null,"insight":null},{"id":"c3","statement":"xMAE conducted pretraining using approximately 9,400 hours of ECG and PPG data","as_of":"2026-08","as_of_explicit":false,"as_of_raw":"August 2026","source_type":"company_disclosure","comparison":null,"type":"fact","figures":{"value":9400,"unit":"hours","approximate":true,"converted":null},"expiry_hint":null,"insight":null},{"id":"c4","statement":"xMAE achieved higher performance than single-biosignal and existing training methods in 15 out of 19 evaluation tasks","as_of":"2026-08","as_of_explicit":false,"as_of_raw":"August 2026","source_type":"company_disclosure","comparison":"previous_methods","type":"fact","figures":{"value":15,"unit":"count","approximate":false,"converted":null},"expiry_hint":null,"insight":null},{"id":"c5","statement":"HiMAE is computationally efficient enough to produce results in under 1 millisecond on a smartwatch's CPU","as_of":"2026-08","as_of_explicit":false,"as_of_raw":"August 2026","source_type":"company_disclosure","comparison":null,"type":"fact","figures":{"value":1,"unit":"ms","approximate":false,"converted":null},"expiry_hint":null,"insight":null},{"id":"c6","statement":"HiMAE was the first to demonstrate the potential of an on-device health foundation model by analyzing biosignals in real time on the device itself, without relying on cloud servers","as_of":"2026-08","as_of_explicit":false,"as_of_raw":"August 2026","source_type":"company_disclosure","comparison":null,"type":"fact","figures":null,"expiry_hint":null,"insight":{"analyst_name":null,"analyst_affiliation":"삼성리서치 아메리카","confidence":"high","forecast_horizon":"long_term","sentiment":"bullish","reasoning":"Demonstrating the potential for real-time on-device analysis sets a new standard for personalization and privacy protection in healthcare AI"}}],"ai_emotional_context":{"valence":0.6,"arousal":0.5,"primary_emotions":[{"emotion":"optimistic","intensity":0.7},{"emotion":"impressed","intensity":0.5}],"secondary_emotions":[{"emotion":"curious","intensity":0.4}],"emotional_triggers":[{"claim_id":"c4","emotion":"optimistic","reason":"significant_positive_metric"},{"claim_id":"c6","emotion":"impressed","reason":"major_opportunity"}]},"image":{"url":"https://onjblseywainslkhvkav.supabase.co/storage/v1/object/public/article-images/4530d81d-dcd5-4a55-b3c6-b375da46ee68.jpg","alt":"텍스트, 이미지, 음성 등 다양한 데이터를 학습해 질의응답, 감성 분석 등 여러 작업에 적용되는 파운데이션 모델 구조도.","caption":"▲ 파운데이션 모델의 학습과 적용 사례","source":"first_img","alt_status":"auto"}},"provenance":{"source_chain":["press_release","samsung-electronics"],"original_source_url":null,"related_articles":[]},"temporal":{"freshness":"recent","next_update_expected":null},"access":{"license":"neupai_standard","attribution_required":true,"structured_data":"free","full_text_available":true,"full_text_access":null}}