Compression Methods¶
compressionKIT implements multiple compression approaches for physiological signals, ranging from classical signal processing to learned neural methods.
Method Comparison¶
| Method | Type | Compression | Latency | Quality | Deployable |
|---|---|---|---|---|---|
| RVQ Autoencoder | Learned | 2×–16× documented | Low | High | INT8 TFLite |
| Wavelet + SPIHT | Classical | 2×–16× | Very Low | Medium | Planned |
| Decimation | Baseline | 2×–16× | Minimal | Low | Trivial |
Design Principles¶
All compression 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 clinical metrics — Not just MSE, but HR/HRV preservation