Voice Trigger to Dispatch Integration Flow
The full technical chain from a user speaking a distress phrase to armed response being dispatched has four distinct layers. Three of these layers are fully solved by existing technology; one requires custom development. This architecture makes the SA voice-activated distress app technically viable without requiring any novel research or unreleased software.
Layer 1: Wake Word Detection (Client-Side)
Picovoice Porcupine runs as an audio processing library within the mobile app’s foreground service. It continuously processes 16 kHz mono PCM audio from the device microphone in frames of approximately 30ms. When the configured wake phrase (e.g. “Help Me”) is detected above the set sensitivity threshold, Porcupine fires a callback to the app’s service layer. This runs entirely on-device with no network dependency — the detection step works offline.
The foreground service must hold a FOREGROUND_SERVICE_TYPE_MICROPHONE permission (Android 14+) and display a persistent notification. On iOS, the app uses UIBackgroundModes: audio with the VoIP entitlement path to maintain microphone access in the background. Both configurations are well-documented and the bSafe global app demonstrates that App Store and Google Play review can be passed with voice-activated SOS.
Layer 2: Alert Construction and API Call (Client-Side)
On receiving the Porcupine detection callback, the app service constructs an alert payload containing: user identifier, current GPS coordinates (from the OS location service), alert type (security or medical), timestamp, and device metadata. This payload is dispatched via HTTPS to the AURA REST API.
AURA’s API is trigger-agnostic: it accepts the same API call regardless of whether the trigger was a button press, a voice detection, a telematics anomaly, or any other signal. The voice-specific complexity is entirely in the client app. From AURA’s dispatch system perspective, a voice-triggered alert is indistinguishable from a manual panic button press. AURA provides sandbox access and API endpoint documentation to integration partners, with a Slack channel for developer support during testing.
Layer 3: Dispatch (AURA Platform)
On receiving the API call, AURA’s backend: (1) identifies the alert coordinates; (2) queries the responder network for the nearest available vetted, PSIRA-registered armed response vehicle; (3) dispatches the nearest unit; (4) opens a monitoring session in the 24/7 AURA control room. The selection and dispatch step happens in “seconds” per AURA’s published documentation. AURA uses webhooks to push real-time status updates back to the triggering app — responder assigned, ETA, location updates.
AURA’s network of 3,000+ vehicles in South Africa (including Fidelity ADT, CSI, and Axon) provides national coverage. Average SA response time is under 10 minutes; best-case in high-density urban areas is approximately 3 minutes.
Layer 4: Connectivity Fallback (Critical)
If the AURA API call fails due to data connectivity loss (no 4G, indoor dead zone, zero prepaid balance), the app must fall back to SMS dispatch. The SMS fallback sends GPS coordinates to the user’s pre-configured trusted contacts and optionally to a control room SMS number. This is the only layer that requires custom engineering beyond off-the-shelf components. SMS delivery works on any voice/data network without requiring active data balance, providing near-universal SA coverage. See Connectivity Failure Risk for the full mitigation architecture.
Integration Timeline Estimate
Based on AURA’s documentation claim of “frictionless integration in days” with sandbox access, a development team integrating for the first time could reasonably expect:
- Discovery and SOP definition with AURA: 1–2 weeks
- Sandbox API integration and webhook testing: 2–3 weeks
- Porcupine wake word engine integration: 2–4 weeks
- Foreground service + battery optimisation handling: 2–3 weeks
- SMS fallback implementation: 1 week
- Testing on device matrix (Samsung, Xiaomi, Huawei, Motorola): 2–4 weeks
- AURA go-live certification: 1–2 weeks
Total estimated: 10–18 weeks for a minimum viable integration with a small team.
Connections
- AURA — core_dispatch_platform, source: https://www.aurasos.com/en-us/technology/integrations
- Picovoice Porcupine — wake_word_engine, source: https://picovoice.ai/docs/faq/porcupine/
- Connectivity Failure Risk — mitigated_by (SMS fallback layer)
- Android Background Voice Recognition — enables (foreground service layer)
- B2B2C Safety App via API Platform — technical_foundation_for
Ontology Voice Trigger to Dispatch Integration Flow [relates] AURA Voice Trigger to Dispatch Integration Flow [relates] Picovoice Porcupine Voice Trigger to Dispatch Integration Flow [relates] Connectivity Failure Risk Voice Trigger to Dispatch Integration Flow [relates] Android Background Voice Recognition Voice Trigger to Dispatch Integration Flow [supports] B2B2C Safety App via API Platform