Technology
Compression built for the shape of a heartbeat.
General-purpose compressors treat a cardiac recording as an anonymous byte stream. We do not. BioRLE-1 exploits the structure that biomedical signals actually have — long quiet stretches punctuated by short, information-dense events.
The problem
The radio, not the sensor, sets the power budget.
Transmission dominates
In a connected monitor, the energy spent putting samples on the air typically dwarfs the energy spent acquiring them. Cutting the payload cuts the dominant term.
Memory is finite
Extended-wear recorders are limited by how much signal they can hold. Compression buys recording days without adding flash.
Fidelity is non-negotiable
Whatever you throw away has to be clinically irrelevant. That is a domain question, not a generic compression question.
The approach
Exploit the redundancy that is actually there.
- Condition the signal Transform the incoming samples so that the predictable, low-information portions of the waveform collapse toward a compact representation.
- Encode the runs A run-length-oriented encoding tuned to biosignal morphology turns those quiet stretches into very few bits, while preserving the sharp events intact.
- Frame the output Emit self-describing frames so a lost or corrupted frame does not poison the rest of the recording — important over lossy wireless links.
- Reconstruct The decoder rebuilds the waveform on the host or in the cloud, from the same open reference implementation your team can audit.
Open source
The core is public. On purpose.
Medical device teams do not adopt compression they cannot inspect, and neither do their regulators. Publishing the algorithm core under Apache 2.0 removes the single largest objection to adopting a compression block from a young supplier.
| Open core | Licensed integration layer | |
|---|---|---|
| Licence | Apache 2.0 | Commercial |
| Contains | Algorithm reference implementation | Hardened RTL, interfaces, collateral |
| Purpose | Inspect, evaluate, verify | Ship it in your product |
| Cost | Free | Licence + volume royalty |
| Support | Community | Direct engineering support |
Repository link to be published alongside the first public release.
Validation
How we know it works.
Functional validation
The encode/decode path is validated against reference vectors, with reconstruction checked sample by sample against the original recording.
Design maturity
BioRLE-1 has completed the validation milestones required to proceed to tape-out. The RTL is stable and the verification collateral travels with the licence.
Your data, your flow
Evaluation runs against your recordings, inside your environment. We do not ask you to take compression numbers on faith.
Traceable by construction
An open reference implementation means the behaviour you verify is the behaviour you can document in your own design history file.
Audit the core before you talk to us.
That is the point of publishing it. When you are ready to put it in silicon, we will be here.