DaLIA-PPG - Heart Rate and Variability Estimation with Multilabel Data and Regression Block - S1_E4

This project demonstrates how to use the PPG-DaLiA dataset for heart rate (HR) and heart rate variability (HRV) estimation using a multilabel regression model. The dataset provides physiological signals, including PPG, ECG, and 3D-accelerometer data, which are processed with the Edge Impulse HR/HRV DSP Block to enable real-time heart rate monitoring and HRV analysis on edge devices.

Key Features of This Project:

• Multilabel Data Handling: Heart rate values are uploaded as continuous labels for training a regression model.
• Regression Block: A Keras-based regression block is used to estimate continuous HR values during inference.
• Motion Compensation: Using both PPG signals and accelerometer data helps to improve HR estimation accuracy by compensating for motion artifacts.
• Edge Impulse DSP: The HR/HRV DSP block extracts key HR/HRV features such as RMSSD, SDNN, and pNN50.

About the Dataset

The PPG-DaLiA dataset contains data from 15 subjects wearing physiological and motion sensors while performing a wide range of daily activities under close-to-real-life conditions. It includes both wrist-worn (Empatica E4) and chest-worn (RespiBAN) devices, capturing multimodal data that combines heart rate ground truth (from ECG) with motion-compensated heart rate estimation (from PPG and accelerometer data).

• Data Characteristics: Multivariate, Time-Series
• Subjects: 15
• Instances: 8,300,000
• Sensors: ECG, PPG (BVP), 3D-accelerometer, EDA, Body Temperature
• Sampling Rates:
  • ECG, respiration, and acceleration from RespiBAN: 700 Hz
  • BVP from Empatica E4: 64 Hz
  • Accelerometer from Empatica E4: 32 Hz
  • EDA and Body Temperature: 4 Hz

The dataset is publicly available for use in regression tasks focused on heart rate estimation and variability analysis, with additional details provided in the dataset's readme file.

Dataset Citation

Citation:

Reiss, A., Indlekofer, I., & Schmidt, P. (2019). PPG-DaLiA [Dataset]. UCI Machine Learning Repository. https://doi.org/10.24432/C53890

For more details and access to the dataset, visit the UCI Machine Learning Repository.