Wearables Design Kit Dataset

This dataset was acquired using the Sensichips SENSIPATCH device and is provided as CSV files. It contains 102 current-measurement features recorded by a photodiode (Photodiode_SFH_2705) from light emitted by six LEDs and reflected by a target surface: 593 nm (Yellow), 645 nm (Red), 950 nm (IR), 523 nm (Green), 468 nm (Blue), and 850 nm (IR). Each LED performs 121 acquisitions, and the photodiode records the in-phase component. The dataset is structured into: TRAINING: measurements collected in a dark-box setup (device covered with black paper, with a small hole exposing LEDs and photodiode). Each run starts with air/dark reference samples and continues with samples acquired after inserting a target (Pantone), enabling ML classification of surface colors and substances. RAW: unprocessed photodiode data for all six LEDs NORMALIZED: RAW data after baseline correction via difference vs. the air/reference measurement TEST: measurements performed in random conditions, intended for model evaluation and iterative improvement. The dataset includes 102 classes/targets: 100 pantones, plus NO-PANTONE (air/dark-box reference) and WHITE-PANTONE (rough white cardboard), with five measurements per target. More information is available at: https://sensichips.com/design-kit/wearables-kit/