A low-cost and portable device for measuring spectrum of light source as a stimulus for the human’s circadian system

Abstract

This study developed a low-cost, portable and wireless spectrometer to measure Spectral Power Distributions (SPD) of light sources with a resolution of 3.5 nm in the visible spectrum using Artificial Neural Networks (ANNs). We conducted the first two experiments to identify the ANNs with the best performance for improving the accuracy of the developed spectrometer. We performed the third experiment under real-world environments to challenge the most accurate ANNs obtained from the previous experiments. We found that the neural network was very effective in reconstructing light source’s SPDs with the known patterns in data showing an error lower than 1% in a controlled laboratory and lower than 18% in the environment in the presence of significant stray light levels. Although the neural network learned the pattern of the electric lighting’s SPDs effectively by an error of 17% in real-world conditions, it had a higher rate of error when it was exposed to daylight with unrecognizable patterns due to the complexity and uniqueness of the SPD. The developed spectrometer offers real-time monitoring of personal lighting exposures. The data is stored on cloud databases using wireless communication, which can be integrated into IoT-based smart lighting systems.