Multi-Parametric Statistical Method for Estimation of Accumulated Fatigue by Sensors in Ordinary Gadgets

TL;DR

This paper introduces a multi-parametric statistical method utilizing sensor data and advanced analysis techniques to estimate and classify fatigue levels in individuals, applicable in everyday scenarios.

Contribution

It presents a novel combination of statistical methods and sensor data analysis for fatigue estimation, validated across diverse individuals and practical for everyday use.

Findings

Fatigue can be effectively estimated using moment, cluster, bootstrapping, and spectral analyses.

The most promising fatigue metric is the distance on the skewness-kurtosis plot.

The method is validated on diverse subjects and can be used by ordinary people.

Abstract

The new method is proposed to monitor the level of currently accumulated fatigue and estimate it by the several statistical methods. The experimental software application was developed and used to get data from sensors (accelerometer, GPS, gyroscope, magnetometer, and camera), conducted experiments, collected data, calculated parameters of their distributions (mean, standard deviation, skewness, kurtosis), and analyzed them by statistical methods (moment analysis, cluster analysis, bootstrapping, periodogram and spectrogram analyses). The hypothesis 1 (physical activity can be estimated and classified by moment and cluster analysis) and hypothesis 2 (fatigue can be estimated by moment analysis, bootstrapping analysis, periodogram, and spectrogram) were proposed and proved. Several "fatigue metrics" were proposed: location, size, shape of clouds of points on bootstrapping plot. The most promising fatigue metrics is the distance from the "rest" state point to the "fatigue" state point (sum of 3 squared non-normal distribution of non-correlated acceleration values) on the skewness-kurtosis plot. These hypotheses were verified on several persons of various age, gender, fitness level and improved standard statistical methods in similar researches. The method can be used in practice for ordinary people in everyday situations (to estimate their fatigue, give tips about it and advice on context-related information).