The use of adaptive Erblet transform with application to ecological data

In this paper, we address the preparation of ecological datasets for data mining. We propose a new adaptive method for automatic dataset construction using Erblet transform, which can be seen as a non-uniform filter bank where the center frequency and the bandwidth of each filter match the ERB (Equivalent Rectangular Bandwidth) scale, followed by data quality assessment using a tonality index (TI). Our ecological database consists of two naturally occurring fish calls produced by the plainfin midshipman fish, Porichthys notatus. The performance of the method is evaluated in terms of K-means clustering on the constructed datasets, and show promising results that would assist in long-term activity monitoring for fish data.