← All work
07Independent zero-to-one initiative2010–2012

Connected safety · IoT · Emergency response

Fall Detection System with GPS

A fall detector was not the product. The product was the complete chain from uncertain motion to a person accepting responsibility and resolving the incident.

IoTGPSEmergency response0→1
RoleProduct Manager
Period2010–2012
StageCustomer discovery · system design · prototype and application development
SystemWearable sensing · RF/base station · GPS/SIM · caregiver applications
System Explainer Portfolio-created model of the connected safety chain · not medical-device UI
Evidence key

Every product visual is labelled by provenance, so original work is never confused with a portfolio-created explanation.

Original Product Artifact Surviving source material from the product work.

A genuine working document, wireframe, screenshot or output that can be shown publicly.

Portfolio Reconstruction Recreated from documented product logic using synthetic or anonymised content.

A recreated product view grounded in the original work, using synthetic or anonymised information.

System Explainer Created for this portfolio to explain product logic; not product UI.

A simplified model created for this portfolio to explain system behaviour, workflow or product logic.

01

The safety problem beyond an emergency button

A conventional emergency button fails when the person cannot press it—and a detected fall still does not prove that help is coming.

Connected safety · IoT · Embedded systems · GPS · Emergency response · Android / iOS

The system had to recognise a potentially serious event, distinguish it from everyday movement, communicate through several technical links, locate the person, notify more than one caregiver and make responsibility visible. The product problem therefore crossed sensing, embedded hardware, connectivity, mobile software and human coordination.

01 User capability

The most serious incident can remove the user from the workflow.

A manual SOS button remained useful, but the product could not depend on the person being conscious, mobile or able to reach it.

02 Detection uncertainty

Normal movement can resemble a fall—and a real fall can be missed.

The product needed a reversible verification path, a cancellation route for false alarms and a test method that compared falls with everyday activity.

03 Technical chain

Sensing, RF, location and cellular delivery could fail independently.

The complete experience depended on multiple devices and communication links. Each degraded condition needed to be visible rather than silently interpreted as safety.

04 Human response

Notification delivery does not establish responsibility.

With several caregivers, the system had to show who had acknowledged the incident, when escalation was required and when the event had actually reached closure.

02

The connected product system

The wearable was one component in a chain that ended with coordinated human action.

The architecture separated sensing, embedded decision logic, local communication, location, outbound alerts and caregiver coordination. Each component had a defined responsibility, but the user experienced them as one promise of safety.

System Explainer Reconstructed from the prototype system and documented workflow
01 Wearable sensing

Accelerometer and gyroscope observe movement and orientation

02 Event logic

Embedded processing evaluates a suspected fall or manual SOS

03 Short-range RF

The wearable transmits the event to the local base station

04 Base station

The powered hub coordinates event, status and outbound communication

05 GPS + cellular

Location context and SIM/SMS delivery support remote response

06 Caregiver network

Android and iOS experiences make acknowledgement and action visible

System ruleEvery link in the safety chain must expose its state. Silence cannot be treated as evidence that the user is safe.
03

Design for detection uncertainty

A movement algorithm needed an alternate trigger, a verification state and a safe route back.

Fall detection was not treated as a binary truth. The system combined inertial signals with manual input and represented uncertainty explicitly before starting the emergency workflow.

System Explainer Portfolio-created state model · not the original firmware algorithm
01ImpactSudden movement event
02OrientationPost-event body position
03StillnessMovement after the event
04Manual inputSOS or false-alarm cancellation
01Monitoring

Everyday movement remains below the event threshold

02Suspected fall

Motion pattern enters a verification window

03Cancelled

The user indicates a false alarm and monitoring resumes

04Confirmed / SOS

Automatic logic or manual input starts the incident workflow

01Detection decision

Use automatic detection alongside manual SOS—and make false alarms reversible.

Automatic detection covered the situation in which the user could not act. Manual SOS covered emergencies that did not look like a fall. A cancellation path acknowledged that false positives were inevitable in a movement-based system and prevented the algorithm from becoming an unquestioned authority.

04

Turn an alert into accountable response

The system had to know whether somebody had actually taken responsibility.

A safety workflow could not end at notification delivery. Acknowledgement, ownership, location, escalation and closure were separate product states because each answered a different operational question.

01Responsibility decision

Track acknowledgement, not only notification delivery.

An outbound SMS or app alert only proves that the system attempted communication. The incident needed a separate acknowledgement state showing that a caregiver had accepted responsibility, followed by response, escalation and closure states.

System Explainer Portfolio-created model of the coordination logic
02Network decision

Use multiple caregivers without creating ambiguous ownership.

A caregiver network reduced dependence on one emergency contact, but it introduced a coordination risk: everyone could assume somebody else was responding. The product therefore needed one visible responder, shared incident status and escalation when nobody accepted the alert.

System Explainer Portfolio-created model of the coordination logic
System Explainer Detection → responsibility → resolution
01Detected

Automatic event or manual SOS

02Verified

False-alarm window closes or SOS bypasses it

03Alerted

Caregiver network receives incident context

04Acknowledged

One caregiver accepts responsibility

05Located

Available location context supports response

06Responding

Help is actively being coordinated

07Escalated

No response or medical need triggers the next path

08Resolved

The incident is closed with an explicit outcome

Product principle

Detection created an incident. Only an explicit outcome closed it.

System Explainer Critical uncertainty remained visible in the workflow
ConditionProduct stateRequired behaviour
01False alarm

Suspected fall

Allow cancellation and return to monitoring
02RF link unavailable

Connectivity degraded

Expose the missing link rather than imply delivery
03Location unavailable

Location degraded

Continue the incident while showing location uncertainty
04No caregiver acknowledgement

Response unowned

Escalate instead of treating notification as completion
05

Prototype validation through controlled contrast

The detection logic was refined by comparing controlled falls with ordinary movement.

The prototype work used a mannequin to reproduce falls from different positions and recorded routine daily motion to understand false-positive patterns. The signals were compared, the decision logic was adjusted and the scenarios were repeated. This was prototype engineering validation—not clinical efficacy testing.

System Explainer Prototype engineering tests · not clinical efficacy evidence
01Controlled fall library
  • Forward / backward events
  • Side and seated transitions
  • Different impact and resting positions
02Everyday movement library
  • Walking and sitting
  • Bending and lying down
  • Routine transitions and device motion
01Record motion
02Compare signatures
03Adjust decision logic
04Retest falls and daily activity
06

Product ownership across connected hardware and applications

I treated the wearable, embedded logic, base station, localisation, communication and caregiver apps as one safety product.

My scope covered user and caregiver discovery, product architecture, incident-state design, hardware and embedded-system coordination, GPS and communication requirements, Android and iOS application oversight, failure scenarios, validation planning and end-to-end prototype delivery. The central PM task was to keep every technical component aligned to the human outcome: a vulnerable person receiving coordinated help.

01

Customer and caregiver discovery around high-risk incident scenarios

02

End-to-end product architecture across wearable, base station, GPS/SIM and apps

03

Automatic and manual triggering, false-alarm and incident-state logic

04

Caregiver acknowledgement, ownership, escalation and closure workflows

05

Cross-functional oversight of embedded hardware and Android/iOS delivery

06

Prototype test scenarios using controlled falls and everyday movement

System Explainer Discovery → connected prototype → validation and refinement
01Discover risk scenarios
02Define connected architecture
03Prototype wearable and base station
04Build incident-state logic
05Develop caregiver applications
06Run fall and daily-motion tests
07Refine thresholds and workflows
08Consolidate prototype system
07

Evidence-bounded outcomes

Delivered work

What the prototype programme made explicit and testable.

01

Completed user and caregiver discovery around falls, remote family support and emergency-response uncertainty.

02

Developed the product-system architecture connecting sensing, embedded hardware, RF, GPS, cellular communication and caregiver applications.

03

Defined automatic and manual triggering, cancellation, acknowledgement, ownership, escalation and resolution logic.

04

Moved the concept into connected-device prototype and application-development work.

05

Established a repeatable prototype-validation method using controlled falls and everyday movement recordings.

Evidence boundary

The work reached customer discovery, detailed product-system design, connected-device prototyping and application-development work. This case does not claim commercial deployment, medical-device approval, clinical efficacy or population-level fall-detection accuracy.

Public evidence note

Public diagrams are date-neutral portfolio reconstructions based on surviving project knowledge and documented product behaviour. They do not reproduce a production medical interface or imply regulatory approval.

08

What the work changed in how I think

A safety product is complete only when uncertainty, responsibility and recovery are designed end to end.

01

Detection is not resolution.

The user problem continues after the algorithm fires. Location, responsibility, response and closure belong in the same product model.

02

False positives are a product problem.

Thresholds matter, but so do cancellation, user trust, caregiver fatigue and the way uncertainty is communicated.

03

Redundancy needs ownership.

Multiple caregivers improve resilience only when one person can visibly accept responsibility and escalation remains explicit.

04

Safety systems must design degraded states.

RF, GPS, cellular delivery and human response can fail independently; the interface must show what is known and what is not.