Pulse-lead Peak Detection

A. Package Installation:

neurokit2 for signal generation and peak labeling

pip install neurokit2 

PyRQA for recurrence plot generation and saving

pip install PyRQA  

tensorflow for CNN model

pip install tensorflow 

PIL for loading recurrence plot as digital data

pip install pillow 

seaborn for heat map (not a core package, only for grid search)

pip install seaborn 

B. Program Introduction

• signal_generation_and_labeling.py

This file shows how to generate ppg signals in bulk and multiple labeling methods. Note that in this project label refers to the probability of being a peak and actual peak information is called peak truth. Illustration examples are also included for reference.

• RP_stuff.py

Codes regarding recurrence plot generation. It starts from examples for signal --> recurrence plot, then gives a function used in final program for saving RP and labels in large quantities. And a data loader function for reading recurrence plots as grayscale data

• adjusted_ppg_simulate.py

An adjusted version of ppg_simulate( ) function I wrote for adding noise, adjusted version is very similar to original one from neurokit2 package except for an extra parameter controlling the degree of noise. And adjusted version outputs 2 signals, one is normal ppg signal and another is noise-added ppg signal.

• Basic_Functions.py

Includes functions for multiple purposes. This is a collection for all functions I wrote,
functions mentioned in other modules are also collected

• feasibility_test.py

An example of creating data and training a model and calculating relevant evaluation standards

• GridSearch.py

The main part of this project, generating data with different noise level and training one model and calculate the distance error matrix & error rate distance

C. Flowsheet