- Mehmet Öztürk
- Celal Can Kaya
- Burak Özdemir
- Fırat Kızılırmak
- Can Dursun Telek
- Alperen Erdem
- Soner Akar
- Derya Kaptan
- Ahmet Koca
The Atracker Smart Wristband empowers you to make each day faster, stronger, and more competitive than the last. The wrist based heart rate monitor on an Android device provides a full fitness scope.
Let's glance at its features.
It is just a regular device, easily turn it on/off. No need to be complex!
At the end of the day, glance at your activities. Check your step counts easily from its application!
Easily accessible batteries. In any hardware problem, possible to easily replace.
Cheaper with regard to other it equivalents.
Though there are negligible issues, it works as intended. Moreover, I like its price!
While developing the project, we were altogether.
This project is based on measuring human activities from an equipment placed on human wrist. The equipment consists of such electronic components as accelerometer and pulse meter. The retrieved data sent through Wi-Fi and Bluetooth channels to both server and android device. Then the data processed on server to store necessary information in database for later use and also the android device processed the data to show the result of stepping and pulse immediately. Furthermore the machine learning algorithm runs on the server to predict whether the user is walking,running,staring up or down etc. As a result, the activities are measured, predicted and demonstrated back to user via android app.
The Machine Learning Algorithm we used is a model of KNN (K Neighbor Classifier). In order train the model, we have collected data during seven days. The collected data is used to train the model. The parameter that the KNN takes is the number of neighbors for the better accuracy. We tried to find the best option that fits our model properly. For that reason, we selected the number of neighbors to be 24. Also, this model is saved for later use which means of prediction.
There are three different screens on the Android application. Initially the logo of our application is demonstrated on the screen and all the permissions are set at the back side of the app for Bluetooth activation and Bluetooth is activated once the app is opened. Then on the different activity screens it tries to create a communication between ESP32 and Android. In order to make this process automatic, we get the address of the ESP32 and we placed it into the code. To check that whether the connection is successful, the Toasty Library is used and we inform the user by sending Toast message.
The implementation of server was completed in C++ using Pistache and RapidJson library for getting post requests and JSON objects. The server has two threads for listening incoming requests and handling jobs. To handle the jobs there are several endpoints called /add, /predict, /getlogs and /change. In order to add the retrieved data to the database an endpoint called /add is created on the server. This endpoint waits an JSON object containing x, y, z axis, gyro values and pulse data.
In the hardware module we used 2 distinct electronic component and 2 onboard components which are accelerometer, pulse meter, Bluetooth and Wi-Fi respectively. Once the board is connected to Bluetooth, it sends the data that are taken from the sensor on the ESP32 to the android device. Also the data is sent to server per two seconds to add to database the necessary information for later use. If it cannot connect to Bluetooth, the data can be retrieved later from the server. In the final version of the device, the SD-card and battery module was inserted. Finally, the whole device is made to work on wrist.