Abstract: Differential optical absorption spectroscopy (DOAS) is widely used for online gas detection in industry. However, 
 when the concentration of industrial gas is low, the spectral absorption is not obvious and the SNR is 
 very low. So if the inversion of industrial gas concentration is carried out by using the traditional 
 methods, it is very difficult to meet the requirements of industrial application. According to the 
 differential absorption spectra of SO$_2$, tritium lamp is used as the light source to collect the 
 high-dimensional data of absorption spectra in 189.73$\sim$644 nm band. And after selecting and preprocessing 
 the absorption spectra data, a deep belief network (DBN) model is established based on the training set data 
 to extract the low-dimensional features of the test data. Furthermore, the extreme learning machine (ELM) 
 is constructed by using the low-dimensional embedding characteristics of training data to realize the 
 calculation of the SO$_2$ concentration. The effectiveness of the proposed model is evaluated, and it 
 seems that the method is more suitable for accurate on-line detection of SO$_2$ concentration in industrial field.

Key words: gas concentration detection, SO2, differential optical absorption spectroscopy, deep belief network, extreme learning machine