PPG (Photoplethysmography)¶

PPG is the most complete signal path in compressionKIT today. The toolkit includes published RVQ bundles, SPIHT/RVQ/hybrid comparison lanes, export artifacts, and a browser demo centered on this signal type.

What is PPG?¶

Photoplethysmography (PPG) measures blood volume changes in the microvascular bed of tissue using an optical sensor. It's the technology behind:

• Pulse oximeters (SpO2 measurement)
• Smartwatch heart rate monitors
• Wearable health trackers

A PPG signal captures the pulsatile component of blood flow, producing a characteristic waveform with systolic peaks and diastolic troughs that repeat with each heartbeat.

Signal Characteristics¶

PPG in compressionKIT¶

For a workflow-level view of what is supported today, see PPG Workflow. For measured 2x-32x RVQ metrics and SPIHT/RVQ/hybrid robustness tradeoffs, see PPG Models.

Data Source¶

The published v1 PPG goldens use the open unified PPG v1 cache built from BIDMC, BUT PPG, PPG-DaLiA, and WESAD. MESA remains a supported restricted source for custom experiments, but it is not used in the published v1 PPG goldens.

uv run python scripts/build_ppg_cache.py \
  --sources bidmc butppg ppg_dalia wesad

Preprocessing Pipeline¶

The PPG preprocessing pipeline uses heliaEDGE augmentation layers:

1. Random crop — Extract a frame_size window from the full segment
2. Layer normalization — Zero-mean, unit-variance per sample
3. Gaussian noise — Additive noise augmentation (training only)

from compressionkit.preprocessing.ppg import build_preprocessor, build_augmenter

preprocessor = build_preprocessor(frame_size=320, epsilon=1e-3)
augmenter = build_augmenter(noise_factor=(0.01, 0.1))

Optional Filtering¶

Bandpass filtering can be applied to inputs and/or targets independently:

data:
  input_filter:
    enabled: true
    low_hz: 0.5
    high_hz: 8.0
    order: 3
  target_filter:
    enabled: true
    low_hz: 0.5
    high_hz: 8.0
    order: 3

Synthetic PPG Generation¶

For data augmentation, compressionKIT can generate synthetic PPG signals via physioKIT:

data:
  synthetic_mix:
    enabled: true
    fraction: 0.1          # 10% synthetic data
    heart_rate_bpm: [50, 120]
    frequency_modulation: [0.1, 0.5]
    ibi_randomness: [0.02, 0.2]

Evaluation Metrics¶

PPG reconstruction quality is assessed using:

• MSE — Mean squared error
• PRD — Percent root-mean-square difference
• Cosine similarity — Waveform shape preservation
• Band-limited metrics — Metrics computed after bandpass filtering to 0.5–8 Hz
• physioKIT alignment — HR, RMSSD, SDNN comparison between original and reconstructed signals