Circadian Rhythms are Not Captured Equal: Exploring Circadian Metrics Extracted by Different Computational Methods from Smartphone Accelerometer and GPS Sensors in Daily Life Tracking

TL;DR

This study compares circadian rhythm metrics derived from smartphone GPS and accelerometer data, revealing significant discrepancies and highlighting that different sensors capture distinct aspects of daily behavior affecting rhythm analysis.

Contribution

The paper systematically analyzes how GPS and accelerometer data produce different circadian metrics, emphasizing the importance of sensor choice in circadian rhythm research.

Findings

Significant differences in circadian patterns between GPS and accelerometer data

Sensor data reveal different sensitivities to device usage and behavior

Circadian metrics vary notably depending on the data type used

Abstract

Circadian rhythm is the natural biological cycle manifested in human daily routines. A regular and stable rhythm is found to be correlated with good physical and mental health. With the wide adoption of mobile and wearable technology, many types of sensor data, such as GPS and actigraphy, provide evidence for researchers to objectively quantify the circadian rhythm of a user and further use these quantified metrics of circadian rhythm to infer the user's health status. Researchers in computer science and psychology have investigated circadian rhythm using various mobile and wearable sensors in ecologically valid human sensing studies, but questions remain whether and how different data types produce different circadian rhythm results when simultaneously used to monitor a user. We hypothesize that different sensor data reveal different aspects of the user's daily behavior, thus producing different circadian rhythm patterns. In this paper we focus on two data types: GPS and accelerometer data from smartphones. We used smartphone data from 225 college student participants and applied four circadian rhythm characterization methods. We found significant and interesting discrepancies in the rhythmic patterns discovered among sensors, which suggests circadian rhythms discovered from different personal tracking sensors have different levels of sensitivity to device usage and aspects of daily behavior.