FallWatch

What it does

FallWatch is a smart wearable that monitors gait, compares real-time movement to thousands of ML-trained fall-risk patterns, and alerts caregivers if instability is detected - preventing falls before they happen.

Your inspiration

Our motivation stemmed from personal experience. Ethan’s grandfather, later diagnosed with Parkinson’s, suffered a fall that marked the start of a sharp physical and emotional decline. Ruchit’s grandmother’s fall had a similar impact - turning her once-active life into one of caution and dependence. We realised subtle gait changes - early warning signs - were often missed, leading to devastating injuries. This wasn’t unique. Conversations with caregivers in elderly homes revealed the same pattern: reactive tools existed, but nothing proactive. The real need was prevention - identifying risk before injury, and that’s what inspired FallWatch.

How it works

FallWatch is a smart insole that detects fall risk before it happens. It contains an ESP8266 chip and motion sensors (accelerometer + gyroscope) that track gait metrics - stride, balance, asymmetry - throughout the day. Every few seconds, the ESP8266 sends a batch of this sensor data via Wi-Fi to a server. The server runs a trained Long Short Term Memory (LSTM) machine learning model, which analyses patterns in the movement data to predict a fall risk score - classified as low, moderate, or high. If a risk is detected, an alert is immediately sent to a connected caregiver app with the user's status, timestamp, and precise location. The system is trained using real and public gait datasets. This architecture enables real-time, personalised fall risk monitoring using affordable, wearable tech.

Design process

Our journey began with a question: why do falls still happen despite available tech? We reviewed WHO data, clinical research, and spoke with caregivers and PhD students. The insight: subtle gait changes occur weeks before a fall - but often go unnoticed. We set out to design a proactive system, not just a device. Our early prototypes explored ankle bands and clip-ons, but user feedback led us to focus on insoles for comfort and day-to-day wear. Each version added new functionality - from a basic IMU logger to streaming real-time data using an ESP8266. The current prototype houses an IMU sensor and ESP8266 in a foam insole. It samples 6-axis motion data, buffers it in 3-second windows, and streams it via Wi-Fi to a server. On the backend, a trained LSTM model (using PhysioNet and test-user data) detects gait instability and returns a fall risk score. If risk is high, the server triggers an alert to a caregiver app. This system - smart insole, real-time ML pipeline, and alert mechanism - has gone through three rounds of iteration. Each time, we refined sensor placement, improved wireless streaming, and optimised the machine learning model to reduce false alerts.

How it is different

Most fall-related devices act after a fall - detecting impact and notifying someone once the person is on the ground. FallWatch takes a different approach. It’s designed to detect the risk of falling in advance. By embedding motion sensors into a simple insole and using an ESP8266 microcontroller, it continuously monitors how a person walks - tracking changes in stride, balance, foot drag, and asymmetry. The data is streamed to a server, where our trained LSTM model detects early signs of instability. If risk is identified, the app not only sends alerts to caregivers or family members but also shares the exact location of the user. What sets FallWatch apart is its focus on early detection, combined with a design that’s practical and affordable. It’s lightweight, discreet, and avoids bulky equipment. At under €30 to produce, it quietly supports daily life - helping prevent falls before they happen.

Future plans

We’re excited to take FallWatch from prototype to real-world impact. Our current working model proves the feasibility of early fall-risk detection through gait analysis using embedded sensors and machine learning. With ongoing input from university professors and the innovation team, we’re refining both the insole and app experience. Next, we’ll test it with our own grandparents, then run pilot trials in elderly homes to gather real feedback. We aim to enhance data accuracy, improve comfort, and begin small-scale production - making FallWatch an affordable, preventive solution for elder care.

Awards