Abstract: The accurate measurement of the composition and the concentration of the toxic and harmful gas in a manned space capsule is one of the key technologies for the spacecraft environmental control and for the life support system. In this paper, In₂O₃ is selected as the gas sensing material to prepare the gas sensors, and the method of the dynamic temperature measurement combined with the neural network model is used to detect the type and the concentration of the gases. The responsive curves of the indium oxide gas sensor for gases of different kinds and different concentrations are obtained. It is shown that after the neural network, by the training model, six different kinds of gases can be identified, and the recognition rate is over 98%. The error for identifying different concentrations of gases is less than 10%, with an average rate of the gradient recognition of more than 95%.