Viva Bem Resistance training biosignals dataset

Here is a multimodal dataset focused on physiological and biomechanical responses during a standardized upper-limb strength-training protocol conducted under controlled experimental conditions. The dataset, referred to as the “Viva Bem Database,” was collected using a Samsung Galaxy Watch 4 (GW4) positioned on the participant’s left wrist, together with synchronized video recordings for experimental documentation, temporal validation, and subsequent biomechanical analysis. The protocol consisted of the isolated and sequential execution of three resistance exercises targeting different upper-body muscle groups: the Pec Deck, focused on the pectoral muscles; the Lat Pull-Down, primarily targeting the latissimus dorsi and associated back musculature; and the Lateral Raise, directed at the lateral head of the deltoid. Exercise sets were interspersed with defined rest periods, while event markers identified the beginning and end of each exercise, set, repetition sequence, and recovery interval. The resulting multimodal structure integrates raw acceleration signals measured in meters per second squared, angular velocity measured in radians per second (rad/s), and continuously acquired heart rate expressed in beats per minute (BPM). These physiological and kinematic time series are synchronized through UNIX timestamps and complemented by video files, typically stored in .mp4 or similar formats, as well as contextual annotations describing exercise type, load, monitored wrist side, repetitions, sets, and rest phases. Events are represented through standardized categorical labels, such as “pec_deck,” “lat_pulldown,” “lateral_raise,” and “rest.” Participant characterization, experimental forms, sensor measurements, and annotated metadata are organized according to a dedicated data dictionary that standardizes variable names, units, technical descriptions, and directory structures across participants and sessions. The data are anonymized and managed in accordance with the Brazilian General Data Protection Law (LGPD). Due to ethical, legal, and contractual restrictions associated with the participation of an industry partner and existing confidentiality agreements, raw data cannot be made publicly available without express authorization from the Principal Investigator and the partner organization. Only non-identifiable aggregate statistics may be openly disclosed. Researchers interested in accessing restricted data must submit a formal request, which will be evaluated on a case-by-case basis and may be subject to confidentiality obligations, attribution requirements, temporary embargoes, and other conditions established by the research partnership. The records will be preserved in secure cloud infrastructure with restricted access and automatic backups, including an additional backup maintained on the Recod.ai server, for a minimum period of five years or longer when required for scientific, legal, or ethical purposes. (2026-06-12)