A project within the module "ubiqutous computing". The aim of the project is to create a system that controls the posture of a user when lifting heavy objects and gives suggestions for improvement. https://youtu.be/2RV6aqpohyI

• Visualize back posture
• Check if person is lifing something
• Recognize a bad posture
• Alert user by beeping or vibrating

• Solder ESP8266 #1
• Solder ESP8266 #2
• Test FSR
• Solder ESP8266 with FSR #3
• make client.ino for fsr
• Make client.ino work
• Make server.ino work
• Test with two clients
• Test with three clients
• Make server.ino work with webserver
• Add status display for clients
• Add beeper to foot https://www.arduino.cc/en/Tutorial/BuiltInExamples/toneMelody
• Add beeper to back
• Request status on boot for clients
• Add function to send data to clients
• Server.js should process data further
• Calibrate MPU6050 Sensors
• Calculate right angles from mpu data
• Display the data on frontend
• Look into neural networks for posture recognition
• Improve webserver data handling
• Smooth values from sensors
• Use multiple axis
• Find better method to time the training
• Restrict training to 20 lines per file
• Fixed null error in training
• Visualize all raw data
• Use visualisations to process data further

• ESP8266/Client -> The sketch for the ESP8266 which have the MPU6050 connected, connect to the AP and send the sensors data
• ESP8266/Server -> The sketch for the ESP8266 which opens the wifi AP and listens for the websocket data
• ESP8266/FSR -> The sketch for the ESP8266 which have the FSRs connected, connect to the AP and send the sensors data
• Server -> The node.js server which get the data from the serial port connected to the Access Point ESP8266
• src -> The parcel webserver files
• data_analysis -> Everything about the data related stuff (mostly csv and jupyter notebooks)

Screenshot of the current version of the frontend

node server/record_data.js

Works best if you have a wireless keyboard/controller and bind a/b to one of the buttons.

1. Connect all the sensors and servermcu to usb
2. Wear the sensors
3. Press 'a' or 'b' to start a recording
4. Perform a motion
5. Press 'a' or 'b' again to stop the recording

Data recording is done via the node js script. Mostly inspired by how charliegerad did it in her project: https://github.com/charliegerard/gestures-ml-js

Main file is data_analysis/complete_set_analysis.ipynb. Takes all files from server/data, parse their names and reads them into a pandas dataframe.

Data analysis is heavily inspired by ATLVHEAD project and his detailed video about it: https://github.com/ATLTVHEAD/Atltvhead-Gesture-Recognition-Bracer