ECG Heart monitor using the STM32 Nucleo Board and TensorFlow Lite

Both the Report and the Presentation explaining the Project in great details are found in the Report_Presentation Folder.

Documentation

Below is the documentation for the most importatnt functions. More documentation is found inside the code itself.
Convention: functionName : parameter1, parameter2, ... : return value1, return value2, ...

Main

• HAL_TIM_PeriodElapsedCallback : void : void

Callback whenever the Timer TIM15 raises an interrupt (when the timer == value AR register) This is where the ADC performs the conversion.

• main : no arguments : returns void (never returns)

Check if ready then performs inference and outputs the result using UART2.

app_x-cube-ai.c

• aiInit : ai_u8 activations : void

Initializes the Neural Network

• aiRun : const void * in_data (input float buffer with the 187 ECG samples), void * out_data (output int buffer of size 5 (5 classes)) : int (prediction status)

Performs the Inference

• MX_CUBE_AI_Init : void : void

Calls aiInit

• MX_CUBE_AI_Process : void : int (the prediction class)

Initializes the input buffer by converting the ECG buffer to bytes then feeding it to aiRun which outputs the prediction then the predicted class is returned