A sparsogram implementation for wildlife intruder detection

Abstract

The main objective of this work is the implementation of a new type of spectral analysis, that is more efficient and that can be more easily read than the spectrogram. This analysis should be used for detecting the intruders from a wildlife environment. It is called sparsogram and its main advantage is that it displays only the important spectral components of a given sound. In this way, one can easily see the dominant frequency bands of different audio signals. Later on, a classification method is implemented based on these frequency measurements on the sparsogram, which assigns the sounds from a natural environment to different classes and subclasses of intruders.