Solution Classification With Portable Smartphone-Based Spectrometer System Under Variant Shooting Conditions by Using Convolutional Neural Network

The rapid development of smartphone-based devices has facilitated on-site measurements, which approach to laboratory-level instrument performance. However, the reliability of these instruments is inevitably limited by the external environment factors. Here, we report on a portable smartphone spectrometer without any external power supply. Using this spectrometer to analyze the diffractive patterns of multiple samples under variant shooting conditions, we verify the influence of external parameters on spectrum. Convolutional neural network is proposed here to perform solution classification under variant shooting conditions and the classification prediction accuracy is more than 99.5%.