The embodiment of the invention provides a voice recognizing method and a voice recognizing device, and relates to the technical field of voice processing. The method is applied to a voice recognizingsystem. The voice recognizing system at least comprises a voice recognizing model, and the voice recognizing model comprises a plurality of voice recognizing sub models. According to the method, aiming at each voice recognizing sub model during each iteration training process, the following operation is carried out till training is finished: the voice recognizing sub models concurrently acquire first voice training data from a training data set; the voice recognizing sub models determine second voice training data according to an acquired voice enhancing parameter and the first voice trainingdata, wherein the voice enhancing parameter is used for increasing background noise for the first voice training data; and the voice recognizing sub models carry out training by adopting the second voice training data, and each voice recognizing sub model is updated according to the training results of the other voice recognizing sub models in the current iteration process. Through on-line generating of the training data, the voice recognizing accuracy is improved.