Categorization of animal sounds using algorithms from diverse applications

The proposed approaches to animal sound characterization are motivated by the philosophical theme of this special session; diverse problems, similar solutions. The literature primarily makes use of some fundamental algorithms. This paper proposes some powerful techniques that have the potential to effect significant advances in animal acoustics. Several classes of algorithms are discussed, along with examples of how they have been used to solve similar problems in diverse applications. The key algorithm areas discussed are (1) statistical feature extraction from two‐dimensional or multi‐dimensional hierarchical transforms, such as wavelet transforms, higher‐order poly‐spectral features, histogram features, texture features, etc.; (2) feature selection algorithms to choose a subset of most useful features and overcome the curse of dimensionality; (3) advanced classification algorithms; (4) performance measures that include estimates of statistical confidence for small sample sizes; and (5) sensor fusion al...