Metro Neighborhood Construction Monitoring System Based on Neural Network

In order to avoid impact of illegal construction and wrong construction along the rail transit, it is necessary to build an environmental monitoring and alarm system. This study uses Mel Frequency Cepstrum Coefficient (MFCC) as the sound feature to build a neural network model to identify the construction sound to monitor whether the construction is in progress in the non construction area. Thanks to the long-distance transmission and low power consumption characteristics of Lora, the identification result is transmitted from the device node to the PC, so the administrator can view the status information of each device node in real time. At last, the accuracy of sound classification and recognition is as high as 99.6%.