An implementation of pre-processing model for PPG, ECG, EEG & Heart Rate signals using Multinominal Logistic Regression Algorithm.

Abstract

In the current lifestyle, smartphone users are rapidly increasing because of the flexibility in the applications available for all kind of users. Those applications provide services that reduce the time and effort of humans in daily activities. It reduces the manual involvement of work because of increased automatic and artificial intelligence-based system understanding modules. The present paper is focused on deriving the impacted factors by the smartphone in terms of physical health and mental health. To measure the stress level of the touch screen users a prediction model is formulated. The basic physiological data of humans are blood pressure, temperature, heart rate, ECG (electrocardiogram), electroencephalogram (EEG), and  photoplethysmogram (PPG). The proposed architecture analyses various physiological data using the embedded sensors and pre-processes the signals using Python Machine Learning. The pre-processing of signals is applied with the help of discrete wavelet transforms to extract the unique peaks present in the signals. The system discusses Haar wavelet function, biorthogonal wavelet function, and Daubechies wavelet commonly called as db4,db8, etc. the collected ECG, EEG, Heart rate, and PPG signals are transformed into pre-processed data through these three kinds of wavelet models. Each model is analyzed here separately to measure the performance of feature extraction. The research work is further extended towards analyzing the frequent patterns of mentioned physic-data with the pre-trained model to measure the accuracy and determine the various class of stress levels.