Abstract
Capturing high-frequency data concerning the condition of complex systems,e.g. by acoustic monitoring, has become increasingly prevalent. Suchhigh-frequency signals typically contain time dependencies ranging overdifferent time scales and different types of cyclic behaviors. Processing suchsignals requires careful feature engineering, particularly the extraction ofmeaningful time-frequency features. This can be time-consuming and theperformance is often dependent on the choice of parameters. To address theselimitations, we propose a deep learning framework for learnable wavelet packettransforms, enabling to learn features automatically from data and optimisethem with respect to the defined objective function. The learned features canbe represented as a spectrogram, containing the important time-frequencyinformation of the dataset. We evaluate the properties and performance of theproposed approach by evaluating its improved spectral leakage and by applyingit to an anomaly detection task for acoustic monitoring.