FILE 02 / WORKING HYPOTHESIS REGISTER

Numbers to test, not performance to advertise.

These values are provisional Norysr design hypotheses. Their literature basis is being re-audited, and none has been measured on Norysr hardware.

Revision 02. Not a prescription, safety envelope or regulatory determination.

00 / CURRENT STATUS

The evidence column is suspended until source verification is complete.

A release review found citation-identity mismatches in the earlier research paper. The public register therefore keeps only the questions and provisional values needed to plan benchtop experiments. It does not claim that the values are established by literature.

Working hypothesis

A proposed value selected for prototype planning. It can change before hardware exists.

Derived budget

Arithmetic assembled from working allocations. It is not a measured timing trace.

Needs specialist review

A safety, clinical or regulatory assumption that the project team is not qualified to determine alone.

Evidence audit pending

No external-source claim is treated as validated until its identifier and supporting text have been checked.

01 / SENSING HYPOTHESES

Start with one shank unit and two muscle channels.

ItemProvisional valueStatus
IMU placementAnteromedial distal shankWorking hypothesis
IMU acquisition6-axis at 100 Hz; compare 200 HzWorking hypothesis
sEMG channels2 channels: tibialis anterior and gastrocnemius medialisWorking hypothesis; placement requires clinician review
sEMG acquisition1 kHz, 16-bit; filtering to be established during bench workWorking hypothesis
Feature windowEvaluate 150–250 ms windowsWorking hypothesis

02 / STIMULATION QUESTIONS

No public value is a safe-use instruction.

Frequency, pulse width, current, electrode placement, command duration and stimulation timing must be defined through electrical-safety work and licensed-clinician review. They are intentionally omitted from this page while the evidence base is re-audited.

Default state

Stimulation off after startup, reset, communication loss, watchdog expiry or sensor fault.

Independent limits

Hardware-level limits should not depend on the application processor or machine-learning output.

Human control

A licensed clinician must define and review any future stimulation envelope before participant-facing work.

03 / INFERENCE HYPOTHESES

The first benchmark is comparative, not clinical.

QuestionProvisional criterionStatus
Can a compact model classify gait phase?Compare against an IMU-only rule baselineExperiment to design
Does sEMG improve timing or robustness?Report participant-level false and missed triggersExperiment to design
Can inference stay local?Profile a quantized model on candidate microcontrollersNo hardware selected
Can calibration help?Compare population and per-user modelsNo Norysr dataset

04 / LATENCY LEDGER

The 23 ms command subtotal remains arithmetic.

StageProvisional allocationStatus
Sensor read≤5 msWorking allocation
Window update and buffering≤10 msWorking allocation
Inference≤5 msWorking allocation
Command transfer≤3 msWorking allocation
Sensor to stimulation command≤23 msUnmeasured derived subtotal

No timing value has been measured on Norysr hardware.

05 / VALIDATION ORDER

Bench evidence comes before participant claims.

  1. Verify sensor synchronization, isolation and default-off behavior.
  2. Measure sensor-to-command latency distributions on candidate hardware.
  3. Compare model output with a transparent rule baseline on an appropriate dataset.
  4. Obtain clinical, ethics, electrical-safety and regulatory review before participant-facing work.