Silent-Caller and Deterioration Detection

Overview#

Knogin Argus detects silent callers and agonal breathing in real time on active emergency calls, alerting the call-taker and supervisor instantly so the correct protocol can be activated without waiting for a manual escalation.

The Silent-Caller and Deterioration Detection module continuously monitors live call audio for clinical and behavioural signals that the caller is in distress, unable to speak, or has lost consciousness. It is designed for ambulance command-and-control and public-safety answering point workflows where the silent-caller pattern is a known indicator of cardiac arrest, stroke, or coercive situations, and where seconds of detection latency translate directly into patient outcome.

Key Features#

• Prolonged Silence Detection: A configurable silence threshold flags a call where the caller has stopped responding, even when the line remains open, so the call is not silently abandoned.

• Agonal Breathing Recognition: Detects the slow, gasping respiration pattern characteristic of cardiac arrest, one of the most clinically important audio signals on a 999 or 112 call.

• Pitch and Breathing Distress Cues: Surfaces shifts in vocal pitch and breathing rate that indicate the caller is struggling to speak or is becoming unresponsive, giving the call-taker earlier warning than silence alone provides.

• Line-Drop Awareness: Distinguishes an active-but-silent caller from a dropped or terminated line, because the correct operational response differs materially between the two states.

• Automatic Supervisor Alerting: When deterioration is detected, a supervisor receives an alert on a dedicated real-time channel and is offered a one-click barge-in to support the call-taker immediately.

• Pre-Emptive Incident Escalation: The platform incident record can have its dispatch priority raised automatically when a deterioration signal fires, without waiting for the call-taker to escalate manually.

• Timeline-Anchored Detection Events: Every deterioration event is written to the unified incident timeline, preserving the precise moment of detection for clinical governance and operational review.

Use Cases#

Cardiac Arrest Recognition#

A caller dials 999, becomes unresponsive, and begins agonal breathing. The platform flags the call as a probable cardiac arrest and prompts the call-taker to activate the relevant resuscitation protocol before the caller can report their own condition.

Coercive or Unsafe-to-Speak Calls#

A caller dials 999 but cannot speak openly. Prolonged silence on an active line triggers a supervisor alert rather than the call being treated as a non-emergency or a nuisance contact.

Caller Collapse Mid-Call#

A caller describing symptoms loses consciousness mid-sentence. The call-taker sees a deterioration banner the moment speech stops, rather than discovering the situation from the audio feed alone.

Supervisor Quality Assurance#

A call-taker handling a complex case receives discreet supervisor backup the moment the audio signals indicate the situation is escalating beyond initial assessment.

Post-Incident Clinical Governance#

Clinical governance teams review the timing between the first deterioration signal, dispatch acknowledgement, and ambulance arrival to evaluate response-time performance and identify improvement opportunities.

Integration#

Knogin Argus exposes deterioration detection as a standard platform capability accessible via REST and real-time channels:

• Call session binding: The detection pipeline runs within the same edge worker that handles the live call audio, so silence and breathing analysis add no additional network round-trip.

• Dispatch assistant surface: Deterioration alerts appear directly alongside dispatch suggestions in the call-taker workstation, so the same panel that presents AMPDS-based priority codes also raises the deterioration banner.

• Supervisor real-time push: Alerts are delivered over the platform's WebSocket supervisor channel. No separate push path is required; any subscriber already listening to call-state events receives deterioration signals in the same stream.

• Incident escalation webhook: When a deterioration signal fires, the platform can automatically raise the incident priority and emit a signed webhook notification to downstream CAD or dispatch systems.

• Incident timeline events: Detection events land on the canonical incident record as structured timeline entries, enabling post-incident review to correlate deterioration timing with dispatch acknowledgement and resource arrival. Events conform to the CloudEvents 1.0 envelope format for consistent downstream consumption.

• Privacy and data residency: All audio analysis runs in-tenant within EU jurisdiction. Raw audio is never exported to external analytics. Only derived detection signals, silence duration, breathing-pattern verdicts, and distress flags, are persisted or transmitted downstream.

Open Standards#

• NENA i3 / NG9-1-1 (NENA-STA-010, NENA-STA-021): incident data and additional data blocks follow NENA Emergency Incident Data Object (EIDO) structures so that deterioration-triggered priority changes interoperate with standards-compliant CAD systems.

• WebRTC (W3C / IETF): voice activity detection uses the WebRTC VAD primitive, running in the same browser-native edge session that carries the live call audio.

• CloudEvents 1.0 (CNCF): deterioration events are emitted as structured CloudEvents envelopes, including silence-detected and agonal-breathing-suspected event types, for consistent consumption by any downstream subscriber.

• Standard Webhooks (HMAC-SHA256): supervisor push notifications and CAD escalation webhooks are signed using the Standard Webhooks specification, giving receivers a verifiable, tamper-evident signal origin.

• AMPDS (Advanced Medical Priority Dispatch System): the dispatch assistant surface maps deterioration signals to AMPDS protocol numbers and determinant codes so call-takers receive a unified priority recommendation consistent with their existing clinical decision support workflow.

• HTTPS / TLS 1.2+: all signal transport between the edge worker, the platform API, the supervisor channel, and call-taker workstations uses encrypted web transport as the baseline security layer.

• SIPREC (RFC 7865 / RFC 7866): call recordings ingested via SIPREC are available to the deterioration pipeline, enabling retrospective analysis and quality-assurance replay alongside real-time detection.

Security and Compliance#

• All audio processing occurs within the tenant boundary and EU data residency zone. Raw audio streams are not forwarded to third-party analytics.
• Supervisor barge-in actions and incident priority changes are written to the immutable audit trail alongside the triggering detection evidence.
• Webhook payloads are signed with HMAC-SHA256 so receiving systems can verify the origin and integrity of every deterioration notification.
• Role-based access controls govern which supervisors receive barge-in alerts and which downstream systems receive escalation webhooks.

Last Reviewed: 2026-05-05 / Last Updated: 2026-05-05