Bioacoustic Signals Denoising Using the Undecimated Discrete Wavelet Transform

Biological populations can be monitored through acoustic signal processing. This approach allows to sense biological populations without a direct interaction between humans and species required. In order to extract relevant acoustic features, signals must be processed through a noise reduction stage in which target data is enhanced for a better analysis. Due to the nature of the biological acoustic signals, the denoising strategy must consider the non-stationarity of the records and minimize the lost of significant information. In this work, a Last Approximation standard deviation algorithm (LAstd) for the processing of bioacoustic signals based on wavelet analysis is presented. The performance of the proposed algorithm is evaluated using a database of owls, which have been modified with different rates of coloured noise. Furthermore, the approach is compared to a standard denoising method from the Matlab Wavelet Toolbox. The results show that the proposed algorithm is able to improve the signal-to-noise ratio of the owl’s registers within a wide frequency range and different noise conditions. Furthermore, the algorithm can be adapted to process different biological species, thus it can be an useful tool for characterizing avian ecosystems.