Most models of respiratory rhythmogenesis are built based on H-H principle, but without any substantive breakthrough.BP network is a widely used model to study artificial intelligent networks.This work is carried out to establish a new model of generation of respiratory rhythm with combination of the 2 models.Five basic respiratory neurons models were built according to H-H style.The data from them were used as input vectors of BP network.Phrenic discharge was considered as output vector.Final weights array was obtained by using MATLAB.Arithmetic platform was designed with the weights array.We chose 5-X-1 BP structure by using reported data about respiratory network models and got the best number (8) of hidden layer neurons.The event of a respiratory cycle was divided successively into 100 sets of data to test the Train, Validation and Test on ratio of 70∶15∶15, respectively.Each R of Train, Validation and Test is bigger than 0.9.It can be concluded that H-H style and BP network can be combined as a new method to build a highly feasible respiratory neural network model.