Denoising Autoencoder Aided Spectrum Reconstruction for Colloidal Quantum Dot Spectrometers

Recently, the colloidal quantum dot spectrometer has received much attention due to its advantages in cost, size, and operation. Yet, just like many other filter-based miniature spectrometers, spectrum reconstruction for the colloidal quantum dot spectrometer is typically prone to the measurement noise due to the correlation of the filters. In this paper, we propose an effective spectrum reconstruction method for the colloidal quantum dot spectrometer, which can recover high-quality spectra in noisy environments. Specifically, we employ a denoising autoencoder, a machine-learning approach, to reduce noise in the filters' raw measurements before performing the reconstruction. After that, we reconstruct the spectra with the denoised data by a sparse recovery algorithm. We investigate the feasibility of the proposed reconstruction approach on a synthetic dataset and an experimental dataset collected by the colloidal quantum dot spectrometer. The results demonstrate that the proposed approach could deliver accurate reconstruction results even when data are corrupted with the measurement noise.