Skip to main content

​
Overview

The EVM Manager provides the main entry point for processing video frames to extract vital signs using Eulerian Video Magnification. It implements an optimized single-pass dual-band approach that is 50-60% faster than traditional methods while maintaining accuracy.

​
process_video_evm_vital_signs

Processes video frames to extract heart rate and respiratory rate using optimized EVM. 
from src.evm.evm_manager import process_video_evm_vital_signs

result = process_video_evm_vital_signs(video_frames, verbose=True)
print(f"Heart Rate: {result['heart_rate']} BPM")
print(f"Respiratory Rate: {result['respiratory_rate']} RPM")

​
Parameters

​
video_frames
list | np.ndarray
required
List of ROI video frames in BGR format. Minimum 30 frames required for accurate analysis.
​
verbose
bool
default:"False"
Enable diagnostic messages and progress information during processing.

​
Returns

​
heart_rate
float | None
Detected heart rate in beats per minute (BPM). Returns None if detection fails or signal quality is insufficient.
​
respiratory_rate
float | None
Detected respiratory rate in breaths per minute (RPM). Returns None if detection fails or signal quality is insufficient.

​
Key Features

​
Optimized Single-Pass Approach

The function implements several optimizations:
  1. Single Pyramid Construction: Builds Laplacian pyramids only once instead of twice
  2. Dual-Band Processing: Processes both HR and RR frequency bands simultaneously
  3. Modular Design: Split into separate modules for maintainability
  4. Performance: Achieves 50-60% speed improvement over traditional implementations

​
Processing Pipeline

​
Frequency Bands

Heart Rate Band

  • Frequency Range: 0.83 - 3.0 Hz
  • BPM Range: 40 - 250 BPM
  • Amplification Factor: 30x
  • Pyramid Level: 3

Respiratory Rate Band

  • Frequency Range: 0.18 - 0.5 Hz
  • RPM Range: 8 - 35 RPM
  • Amplification Factor: 50x
  • Pyramid Level: 2

​
Usage Example

Complete example from experimental validation: 
import cv2
from src.face_detector.manager import FaceDetector
from src.evm.evm_manager import process_video_evm_vital_signs
from src.config import TARGET_ROI_SIZE

# Initialize face detector
face_detector = FaceDetector(model_type='mediapipe')
frame_buffer = []
BUFFER_SIZE = 200

# Capture and buffer frames
cap = cv2.VideoCapture('video.mp4')
while len(frame_buffer) < BUFFER_SIZE:
    ret, frame = cap.read()
    if not ret:
        break
    
    # Detect face and extract ROI
    roi = face_detector.detect_face(frame)
    if roi:
        x, y, w, h = roi
        roi_frame = frame[y:y+h, x:x+w]
        roi_frame = cv2.resize(roi_frame, TARGET_ROI_SIZE)
        frame_buffer.append(roi_frame)

# Process with EVM
result = process_video_evm_vital_signs(frame_buffer, verbose=True)

if result['heart_rate']:
    print(f"Heart Rate: {result['heart_rate']:.1f} BPM")
if result['respiratory_rate']:
    print(f"Respiratory Rate: {result['respiratory_rate']:.1f} RPM")

cap.release()
face_detector.close()

​
Error Handling

The function handles various error conditions gracefully:
  • Insufficient Frames: Returns None values if fewer than 30 frames provided
  • Invalid Input: Validates input type and returns None for invalid data
  • Processing Errors: Catches exceptions and returns None with optional verbose logging

​
Performance

​
Benchmark Results

Tested on typical video processing scenarios:
  • Processing Speed: ~25-30 FPS on Raspberry Pi 4
  • Time per Frame: ~30-40ms average
  • Buffer Processing: ~6-8 seconds for 200 frames
  • Accuracy: MAE < 5 BPM for heart rate

​
Configuration

All frequency bands and thresholds are configured in src/config.py: 
# Heart Rate Parameters
ALPHA_HR = 30        # Amplification factor
LOW_HEART = 0.83     # Low frequency (Hz)
HIGH_HEART = 3.0     # High frequency (Hz)
MIN_HEART_BPM = 40   # Minimum BPM
MAX_HEART_BPM = 250  # Maximum BPM

# Respiratory Rate Parameters
ALPHA_RR = 50        # Amplification factor
LOW_RESP = 0.18      # Low frequency (Hz)
HIGH_RESP = 0.5      # High frequency (Hz)
MIN_RESP_BPM = 8     # Minimum RPM
MAX_RESP_BPM = 35    # Maximum RPM

Build docs developers (and LLMs) love

Get started for freeTalk to us