Camera-Based Heart Rate and Heart Rate Variability Measurement (CBHRM)

Overview

This project is part of a Bachelor's thesis focused on real-time heart rate and variability detection using camera-based methods. It aims to provide a non-invasive way to measure heart rate and heart rate variability (HRV) using image processing and signal analysis techniques. Application based on proposed pipeline[1]

Important: OpenCV needs to be build from source with Gstreamer

Features

• Heart Rate Monitoring: Utilizes camera input to detect and monitor heart rate.

• Heart Rate Variability Analysis: Offers HRV analysis using advanced signal processing techniques.

• Non-Invasive Method: Employs a camera-based approach, eliminating the need for physical contact.

• Configurable Settings: Includes various settings for camera resolution, filtering parameters, and more, as defined in settings.toml.

• Interactable Dashboard: - Dashboard for recording and visualising various signals inclusing:

  • HR (with or without reference device)
  • HRV (with or without reference device)
  • rPPG Signal
  • Post Processed rPPG signal
  • Correlation Plots
  • ...

• Reference Device evaluation: utilises Openant to read in ANT+ compatible chest strap

• Evaluation of Accuracy: Evaluates the reference device HR against the estimated HR

• Reference Device HRV calulation: Caluculates the HRV of a ANT+ compatible chest strap

Installation

1. Clone the repository:

   git clone https://github.com/parisj/CBHRM.git

2. Install required dependencies:

   pip install -r requirements.txt

Usage

To run the application, execute:

cd CBHRM
python run_application.py

Configuration Settings

Camera Settings

• fps_camera: The frame rate of the camera in frames per second. Default is 20.
• resolution: The resolution of the camera. Default is 960x720 pixels.

Video Source Settings

• live: Determines if the video source is live (true) or pre-recorded (false). Default is false.
• path: The file path to the pre-recorded video. Default is "dataset/P1/Talking/recording_3.avi".
• evaluate_dataset: A boolean flag to indicate whether to evaluate the dataset. Default is true.
• path_measurements: File path to the dataset measurements. Default is "dataset/P1/Talking/recording_3.csv".

  For using the live webcam stream:

  live = true
  evaluate_dataset = false

  for using prerecorded video:

  live = false
  evaluate_dataset = false

  for evaluating a prerecorded dataset

  live = false
  evaluate_dataset = true

Filtering Parameters

• fs: Sampling frequency. Default is 20.
• lowpass_order: Order of the lowpass filter. Default is 6.
• wn_lowpass: Normalized cutoff frequency for the lowpass filter. Default is 0.6.
• pos_window_l: Position window length. Default is 32.
• cutoff_window: Cutoff window frequencies. Default is [0.235, -0.235].
• bandpass_order: Order of the bandpass filter. Default is 25.

Heart Rate Measurement Settings

• time_window: Time window for measurement in milliseconds. Default is 200.
• start_delay_peak_detection: Delay before starting peak detection in milliseconds. Default is 15.
• distance: Minimum distance between peaks. Default is 6.
• prominence: Minimum prominence of peaks. Default is 0.

Evaluation Parameters

• delay: Delay in evaluation in samples. Default is 85.
• len_hr_min: Minimum length for heart rate data. Default is 21.
• fps: Frames per second for evaluation. Default is 20.
• dashboard_refresh_time: Dashboard refresh rate in milliseconds. Default is 1000.
• min_ref_measurements: Minimum reference measurements. Default is 500.
• calibration_time_ref: Calibration time for reference in samples. Default is 300.
• min_ref_hrv: Minimum reference for HRV. Default is 90.
• calibration_time_ref_hrv: Calibration time for HRV reference in samples. Default is 50.

Result Configuration

• path: Path to save results. Default is "dataset/P2/breathing2.csv".
• write: Enable writing results. Default is true.

Reference Measurement Settings

• hrv_window: HRV measurement window in samples. Default is 200.
• device_id: Device ID for the reference measurement. Default is 20074.

Record Data Set Settings

• fps_camera: Frames per second for the camera in the dataset recording. Default is 20.
• resolution: Camera resolution for the dataset recording. Default is [960,720].
• device_id: Device ID for the dataset recording. Default is 20074.
• video_path: Path to the video file for the dataset recording. Default is "dataset/P2/breathing2.avi".
• readings_path: Path to the readings file for the dataset recording. Default is "dataset/P2/breathing2.csv".

Add new reference device

1. Scan with openant for devices and write json will found devices

openant scan --outfile devices.json

2. Extract device ID

3. Replace device ID with new ID

   Tip: Duplicate settings.toml and name old one: settings_old_device_ID.toml

Folder Strucutre

CBHRM/
    .gitignore
    README.md
    requirements.txt
    run_application.py
    settings.toml
    dataset/
        P1/
            Post_Workout/
            Rotation/
            Steady/
            Talking/
        P2/

            Rotation/
            Steady/
            Talking/
        P3/
            Post_Workout/
            Rotation/
            Steady/
            Talking/
        P4/
            result.csv
            Post_Workout/
            Rotation/
            Steady/
            Talking/
        Plots/
    results/
    src/
        blackboard.py
        control.py
        dashboard.py
        heart_rate_monitor.py
        image_processor.py
        signal_processor.py
        video_stream.py
        __init__.py
        devices/
            devices.json
        util/
            heart_rate.py
            histogram_inspection.py
            legacy_dashboard.py
            livefilter.py
            load_pkl.py
            plot_frequency.py
            plot_functions.py
            recoord_dataset.py

---

[1] A. Gudi, M. Bittner, R. Lochmans, and J. van Gemert, ‘Efficient Real-Time Camera Based Estimation of Heart Rate and Its Variability’. arXiv, Sep. 03, 2019. Accessed: Dec. 10, 2023. [Online]. Available: http://arxiv.org/abs/1909.01206