A machine-learning assisted measurement device for circadian lighting based on spectral sensors

Abstract

Light has an undeniable impact on the human body, as it can to some extent affect hormone secretion and emotional changes. Spectral power distribution (SPD) is the main indicator for evaluating the quality of light sources, but traditional spectral measurement equipment is bulky and expensive, and cannot be widely used in our daily life. In order to fill this gap, this article designs a low-cost and small lighting measurement device for measuring the circadian lighting, which obtains spectral data from 8 channels in the visible light range through multi-channel spectral sensors. Machine learning methods are used to reconstruct the SPD of 81 wavelength data points, thereby improving the accuracy of designed measurement device. This device can simultaneously achieve real-time measurement of SPD and real-time monitoring of circadian related parameters, and return circadian related parameters (such as circadian action factor, melanopic efficacy of luminous radiation, equivalent melanopic lux, etc.). Results have found that the error of circadian parameters measured by this equipment is less than 5%.