Abstract: Efficient use of construction vehicles is an effective way to save costs in engineering project management. Working condition recognition of construction vehicles in an unsupervised environment is an effective means to achieve efficient use of construction vehicles. At present, vehicle intelligence with GPS and other technologies as the core management system do not recognize the working condition of the construction vehicle. This paper proposes a working condition recognition method for construction vehicles based on GRU re-current neural network. By analyzing the audio signals generated by construction vehicles under different working conditions, the Mel Frequency Cepstral Coefficient is extracted as the main features and the GRU recurrent neural network model is constructed. Perform training and recognition. Experiments show that this method can realize effective recognition of working conditions of construction vehicles..

Key words: engineering vehicle, working condition recognition, audio feature, Mel frequency cepstral coefficient, recurrent neural network