The invention provides an expressway traffic flow state identification method based on a deep neural network and relates to the intelligent traffic technology field. The method comprises the followingsteps of classifying and defining a traffic flow state, carrying out noise reduction processing and characteristic extraction on an audio signal, using DNN to carry out modeling so as to acquire a DNN model identifying an expressway traffic flow state, and pre-training the DNN model; then, tuning a parameter of the DNN model; using a hidden Markov model HMM to decode an expressway traffic flow state identification model; and finally, using the DNN model to estimate observation probabilities of the audio signals of different expressway traffic flow states, and giving an identification result of the expressway traffic flow states according to the calculated probability. When an existing image analysis technology is used to detect traffic information, image analysis accuracy is poor, a calculated amount of dynamic image analysis is large and so on. By using the method of invention, the above problems are solved.