Smartphone application for automatic classification of environmental sound

Sounds are an important source for events in the environment. They convey information about events even when these are not in line of sight. This information can warn a person that a danger is occurring. E.g. a pedestrian can estimate the distance of a vehicle and if the vehicle is approaching or departing. In this contribution a mobile sound classification system based on a smartphone running the Android operating system is presented. The software was developed in Java and applies pattern recognition methods to recognize environmental sounds. It extracts thirteen Mel Frequency Cepstral Coefficients (MFCC) from the data collected by the microphone and classifies the sound using the neural network. The use of pattern recognition methods makes the approach easily adaptable to different sounds. As application example the presented system is trained to recognize sounds of emergency vehicles in road traffic.