Research On Alarm Of Train-Approaching Based On The Technology Of Embedded Audio Recognition

A method of train audio recognition is studied in this paper. Audio sensor can hear the sound of a faraway train. This method improves the distance of alarm of train-approaching and security.The key of this method is that the sound of train can be effectively recognized. Because the sound of the train can be easily interfered by other noise such as the noise of the knock on the rail, the noise of other vehicles, and the noise in the environment. If not to deal with the noise, the result is invalid. The veracity of train's audio recognition depends on the research of characteristic model of the train's sound and the exactitude of the audio recognition algorithm.The train's audio recognition algorithm is designed in this paper. The time domain correlation characteristic and line-spectrum characteristic in frequency domain of samples have been analyzed in this paper. This task establishes the templates of train's sound and simulates on the platform of MATLAB by the use of time domain correlation characteristic and line-spectrum characteristic matching of frequency domain from acoustics recognition angle. At the same time, we confirm the audio recognition algorithm reliable by lots of the sound samples. By analyzing the exactitude and computing time, summarizing the merits and limits of algorithms, an improved method is proposed.This paper designs a solution of the train's audio recognition by setting up a hardware simulation platform based on the DSP system. The content of research includes hardware circuit design, system configuration and sound acquisition programs. The result of the experiment indicates that this design can accurately recognize the sound of the coming train by limited sound templates.