An application for wearable technology tracking activity of persons engaged in occupational therapy allows for unattended accounting. Multiple accelerometers and gyroscopes capture g’s as patients perform activities such as cycling or step climbing. To accurately account for the duration of each activity performed by each patient, a system of recognizing the pattern of movement data is needed.
Wearable sensors transmit data through wireless radio in to PubNub’s IoT Data Broker. This data is currently forwarded to AWS to be archived for the long term although there are no clear uses of the fine grained data.
The wearable sensor data is piped directly to Falkonry from PubNub. Falkonry simultaneously examines patterns in the high frequency movement sensors to recognize the exact physical activity being performed. Using this AI model, the duration of patient activity can be quickly and accurately classified — without the need for data scientists and software programming. Falkonry continues to classify the patient activity returning real-time classification to Amazon Kinesis.
By using Falkonry, human activity monitoring application providers are able to build a library of classified activity based on accelerometer signal patterns. Falkonry receives continuous feeds of accelerometer signal data and returns the derived human activity (classification) to the application. As new activities are added, new patterns of activity are introduced. Falkonry’s continuous learning feature makes recognition and addition of new activities quick and easy.