Spectral Analysis of Array of Sensors Based on Hierarchical 1D-DCNN

Abstract

The research focus of this research will construct a novel 1D-CNN and hierarchical neural network architecture, namely Hierarchical Deep Neural Network (HDNN), and apply it to the development of spectral oxygen concentration measurement technology (DSL). Regression algorithm based on neural network. The HDNN we designed can segment the spectral bands in visible light and near-infrared to improve the measurement accuracy. The average error value in the training set is only 0.01091%, and the average error value in the test set is only 0.01117%, which is much smaller than the 2% of the traditional measurement deviation in the past. Demonstrate that this innovative approach has greater precision. All these analysis information shows that our method based on 1D-CNN and HDNN model for multispectral measurement is potential and highly feasible.