【摘要】 为了有效监测室内有害气体,提出了一种改进的深度残差网络的室内有害气体定量分析方法。该方法使用残差网络来提高深度学习方法学习干扰信号特征的能力,并提高混合气体定量分析的准确性以及模型的鲁棒性。将软阈值化用作非线性层,并将其嵌入到深度神经网络中,以消除非必要的噪声特征。该方法结合了深度学习与传统信号处理降噪算法的优点,通过使用具有不同干扰水平的甲醛气体样本集进行了验证,实验结果表明该模型在有干扰的场景下仍具有较高的准确率。更多还原

【Abstract】 In order to effectively monitor indoor harmful gases, this paper proposes an improved deep residual network method for quantitative analysis of indoor harmful gases. This method uses residual networks to improve the ability of deep learning methods to learn the characteristics of interference signals, and improve the accuracy of quantitative analysis of mixed gases and the robustness of the model. The soft threshold is used as a non-linear layer and embedded in the deep neural network to eliminate unnecessary noise features. This method combined the advantages of deep learning and traditional signal processing noise reduction algorithms, and was verified by using formaldehyde gas sample sets with differ-ent interference levels. The experimental results show that the model is still highly accurate in scenes with interference.更多还原