When it comes to fitness tracking, the average consumer would most likely reach for a smartwatch or similar wearable band. These all work by using their internal accelerometers and gyroscopes to sense motion and detect when a certain action, such as stepping or lifting a weight, has been completed. But to further simplify the process by eliminating the need to select a workout before starting an exercise routine, Nekhil and Shebin Jacob have worked together to come up with the GetFit fitness tracker.
The GetFit is a battery-powered device that uses machine learning to detect not only when an action has been done, but also what kind of workout is being performed. They achieved this by gathering plenty of samples from a Nano 33 BLE Sense’s onboard accelerometer and training a Keras model with the help of the Edge Impulse Studio. It can accurately identify between arm circles, pushups, squats, and anything else in the future while also disregarding anomalous data.
The Arduino sends the now-recognized motion to a connected smartphone over Bluetooth® Low Energy where it’s then used to calculate the number of calories burned and display weekly activity levels. Best of all, this data is tied to an account in a Firebase database for easy transferability.
To read more about GetFit, you can read the duo’s write-up here on Hackster.io.