PPG Signals with Motion Artifacts from Smart Cane Sensors for Heart Rate Estimation

This dataset accompanies the publication: Villalba-Bravo, R., Ruiz-Barroso, P., Castro, F. M., Trujillo-León, A., Guil Mata, N., & Vidal-Verdú, F.Deep Neural Network to Remove Motion Artifacts from Heart Rate Sensor Embedded on Handle Cane.IEEE SENSORS 2024, pp. 1–4.View Publication Description This dataset contains physiological data collected from 17 participants who walked on a treadmill during an acoustic startle-response experiment. The goal was to study the effect of motion artifacts on PPG signals collected from a smart cane and compare them to a clean, stationary reference. Each participant's folder includes: Heartbeat counts over time windows (beats_number.txt) Motion-contaminated PPG signal from the smart cane (HP_LP_baston_series.txt) Clean PPG reference signal from a stationary sensor (HP_LP_ref_series.txt) Folder Structure S_01/├── beats_number.txt├── HP_LP_baston_series.txt├── HP_LP_ref_series.txt...S_17/ File Descriptions 1. beats_number.txt Format: CSV (3 columns, no headers) Content: Number of systolic peaks counted in the reference PPG signal within a sliding window. Columns: window_start_time: Start time of the window (Unix timestamp, e.g., 1718353973.77746) window_end_time: End time of the window systolic_peaks_count: Number of detected peaks 2. HP_LP_baston_series.txt Format: CSV (2 columns, no headers) Content: Raw PPG signal acquired from the handle of a smart cane. This signal contains motion artifacts due to walking. Columns: ppg_value: Contaminated PPG signal (0.80 Hz - 3.33 Hz FIR bandpass applied) timestamp: Unix timestamp of each sample (e.g., 1718353970.03125) 3. HP_LP_ref_series.txt Format: CSV (2 columns, no headers) Content: Reference PPG signal recorded from a stationary, fixed sensor, used to reliably detect heartbeat peaks. Columns: ppg_value: Reference PPG signal (0.80 Hz - 3.33 Hz FIR bandpass applied) timestamp: Unix timestamp of each sample Additional Information All time formats are in Unix POSIX timestamp with fractional seconds. Heartbeat peaks were counted using the reference signal to validate motion artifact removal models. Citation If you use this dataset, please cite the original article: R. Villalba-Bravo, P. Ruiz-Barroso, F. M. Castro, A. Trujillo-León, N. Guil and F. Vidal-Verdú,"Deep Neural Network to Remove Motion Artifacts from Heart Rate Sensor Embedded on Handle Cane,"2024 IEEE SENSORS, Kobe, Japan, 2024, pp. 1-4,doi: 10.1109/SENSORS60989.2024.10784567.