DEV Community

Cristhiam Leonardo Hernández Quiñonez
Cristhiam Leonardo Hernández Quiñonez

Posted on Edited on

SmartSpectra Edge Sentinel – A Consent-First Android Pulse Demo

Presage SmartSpectra SDK Challenge Submission

HormigasAIS Logo

This is a submission for the Presage SmartSpectra SDK Challenge.

What I Built

SmartSpectra Edge Sentinel is a small Android prototype that uses Presage SmartSpectra 3.4.0 to display a live pulse value obtained through the phone’s front camera.

Before any measurement starts, the app shows an in-app consent dialog. The participant can cancel or continue to the Android camera-permission check. The app also provides clear Start and Stop controls.

I developed and tested the build on a Samsung Galaxy A16 using Termux. The goal was to make the live-signal flow concrete and inspectable on a real phone, not to create a clinical product.

Demo

The unlisted YouTube video shows:

  • App launch
  • In-app consent screen
  • Camera-based measurement flow
  • Pulse values appearing and changing during the session

👉 Watch the demo directly on YouTube

Code

Public source repository (includes build instructions and interoperability notes):

smart-spectra-hormigas-edge

The repository does not distribute an APK or AAB. A development API key injected into a compiled package can be extracted, so builds that contain an active key are kept local.

How It Responds to Human Signals

The app observes SmartSpectra metrics and displays the latest available value from cardio.pulseRateList as pulse in beats per minute.

The demo shows an initial waiting period before a value appears, followed by changing pulse readings while the session is active.

Current implementation and boundaries:

  • The Android callback displays live pulse metrics evaluated over a candidate session window via SmartSpectraConsensusEvaluator.
  • When stability criteria are met (stable=true, positive confidence, increasing timestamps), local data is channeled through PresageLbhBridge and LbhEncoder to construct a validated 16-byte fixed LBH frame strictly in volatile memory.
  • The app sets imageOutputEnabled = false and does not deliberately write biometric fields to disk or transmit LBH frames over a network.
  • The repository also contains separate Python laboratory tests for the public lbh-sdk packet interface and an experimental fixed-length 16-byte representation.

Prize Categories

Best Thoughtful Real-World Use

This prototype focuses on an explicit, visible camera-measurement flow and a clear display of the pulse signal, with strong emphasis on consent and transparent data handling. It is an engineering demonstration, not a diagnostic, treatment, or medical-decision tool.

Consent and Data Handling

  • The app presents its own consent dialog before measurement.
  • It checks/requests Android camera permission before starting the SmartSpectra session.
  • Participants can cancel before starting and can stop an active measurement at any time.
  • The application sets imageOutputEnabled = false and does not deliberately record or export camera video, persist biometric measurements, or transmit measurements to an external endpoint; volatile 16-byte LBH frames are cleared after encoding.
  • The project has not independently inspected SDK network traffic or verified provider-side retention/deletion practices.

Presage documents SmartSpectra session telemetry as enabled by default. Its documentation describes aggregate diagnostic telemetry and states that it excludes raw video/image frames and measured vital values. This app does not explicitly disable telemetry. These are provider statements, not the result of an independent network audit.

See the project’s consent and data handling notice and Presage’s telemetry and privacy documentation.

This prototype is not a medical device and is not intended for diagnosis, treatment, or medical decision-making.

Top comments (1)

Collapse
 
thrumanshow profile image
Cristhiam Leonardo Hernández Quiñonez • • Edited
  1. Telemetry & Consent: Runtime state and vendor diagnostics are fully transparent; biometric payloads are strictly restricted to volatile RAM buffers and never touch disk or external endpoints.

​2. Deterministic 16-Byte LBH Framing: SmartSpectraConsensusEvaluator and PresageLbhBridge/LbhEncoder normalize validated consensus windows into a fixed-width binary packet, verifiable instantly via verify_lbh_vector.py (canonical vector 0xAE... passing cleanly).

​3. ARM64 / Termux Reproducibility: To prevent hardcoded dev key exposure, builds use modular Gradle setups via local.properties rather than bundling vulnerable prebuilt APKs, keeping edge execution fully sovereign.

​4. Handling Mobile Jitter: Sliding candidate windows enforce strict stability gates (SDK stable flag + increasing timestamps) to trade initial warm-up latency for clean, jitter-free output.
​Status: Fully verified, code-backed, and closed. No architectural gaps between documentation and implementation.