A fitness tracker built using an Arduino Nano 33 BLE Sense. Tracks the user’s step count, walking time, sitting time and standing time.
Technologies used
- TensorFlow Lite for Microcontrollers
- P5.js
- p5.ble.js
- Chart.js
- HTML, CSS, JavaScript
Project Enclosure
The above images show the enclosure for the project. A clip was attached to the back of the enclosure using epoxy. The Arduino board is powered by a rechargeable 9V battery.
User Interface
The user interface that shows the activity information.
Project Build Process
- The main goal of the project was to build a prototype fitness tracker that used a machine learning model to classify between different activities- walking, sitting, standing.
- The first step involved collecting data to train the model. For this I used the Tiny Motion Trainer by Google Creative Lab. Following the instructions given on their website, I connected to their online interface and used the following settings:
- Capturing threshold: 0.08
- Number of samples: 50
- Delay between captures: 0.2s
- I captured 30 data samples for each of walking, sitting, and standing. I then downloaded the csv files to use for training the model.
- I trained my model on Google Colab using a Python script I wrote (Links to the code given at end of article).
- Once the model was trained, I downloaded the model.h file and put it in the same folder as the Arduino sketch.
- After the above steps, I was left with the final task of building the user interface. I decided to use standard web technologies and P5.js as I’m comfortable with using these technologies. The whole project took me a month to build and it was an amazing learning experience.
Links
~ Ushan Fernando😀💻