Indoor-Outdoor Detection Using Head-Mounted Lightweight Sensors

Spending adequate amount of time outdoor is helpful for both physical and psychological health. Indoor-outdoor (IO) detection provides useful information for end users such as the inside and outside time spent monitoring. In addition, IO detection is extremely important in IO navigation and localization. Several authors focused on the IO detection using the smartphone sensors as well as the light sensors available in selected wearable devices. The aim of this work is to compare the accuracy of different machine learning algorithms in discriminating IO environments using a new generation of color light sensor mounted on the head, both standalone and in combination with other lightweight sensors, i.e., UV sensor, pressure sensor, accelerometer, and gyroscope. Data have been acquired in different days on a population of 28 subjects. Six machine learning algorithms have been tested on the overall acquired dataset. Among the tested algorithms, bagged trees and naive Bayes showed the best performances in terms of accuracy, respectively, around 87 and 89% involving both color and UV sensors. Furthermore, the naive Bayes algorithm showed the higher performances in critical environments such as semi-indoor and semi-outdoor ones.