Compression Methods¶
compressionKIT compares multiple codec families through the same release and scorecard surfaces. The v1 HuggingFace bundles are RVQ neural codecs; SPIHT and hybrid lanes are registered locally so clean-signal and noisy-wearable tradeoffs can be measured in a standardized way.
Method Comparison¶
| Method | Type | Current role | Deployment status |
|---|---|---|---|
| RVQ Autoencoder | Learned neural codec | Published v1 PPG/ECG bundles and primary runtime path | INT8 LiteRT/TFLite encoder + codebook artifacts |
| Wavelet + SPIHT | Classical DSP codec | Clean-signal and faithfulness baseline; registered golden comparison lane | Local deploy package support; not in v1 HuggingFace bundles |
| Hybrid + SPIHT | Learned/DSP hybrid | Wearable-noise and artifact comparison lane | Local deploy package support; not in v1 HuggingFace bundles |
| Decimation | Simple baseline | Sanity baseline for compression-ratio studies | Trivial local implementation |
How to choose what to inspect first¶
| Question | Best starting page |
|---|---|
| What can I download today? | Model Zoo |
| How do PPG methods behave under empirical noise or motion? | PPG Models |
| How do ECG methods behave under SNR and contact-artifact sweeps? | ECG Models |
| What are the CR-vs-fidelity numbers? | PPG CR vs Fidelity, ECG CR vs Fidelity |
| What do the metrics mean? | Validation Scorecard |
Design principles¶
All release-facing methods in compressionKIT follow these principles:
- Portable to embedded C — No dynamic allocation, fixed memory layouts
- Quantization-friendly — Only operators that work with INT8/INT16x8 quantization
- Configurable via YAML — Major parameters exposed through configuration files
- Evaluated on signal utility — Not just MSE, but modality-specific metrics and noise/artifact behavior
CR vs. Fidelity Decision Artefacts¶
Summary tables of the v1 RVQ goldens with codec compression ratios, optional codec+prior effective compression ratios where available, and noise-aware fidelity metrics (PRD, PRDN-noise, HR MAE, QRS- / pulse-band PSD error, coherence, stitching seam ratio):
Regenerate from existing scorecards with python scripts/build_cr_vs_fidelity.py --modality {ecg,ppg}.