English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Varying conditions between the data seen at training and at application timeremain a major challenge for machine learning. We study this problem in thecontext of Acoustic Scene Classification (ASC) with mismatching recordingdevices. Previous works successfully employed frequency-wise normalization ofinputs and hidden layer activations in convolutional neural networks to reducethe recording device discrepancy. The main objective of this work was to adoptfrequency-wise normalization for Audio Spectrogram Transformers (ASTs), whichhave recently become the dominant model architecture in ASC. To this end, wefirst investigate how recording device characteristics are encoded in thehidden layer activations of ASTs. We find that recording device information isinitially encoded in the frequency dimension; however, after the firstself-attention block, it is largely transformed into the token dimension. Basedon this observation, we conjecture that suppressing recording devicecharacteristics in the input spectrogram is the most effective. We propose afrequency-centering operation for spectrograms that improves the ASCperformance on unseen recording devices on average by up to 18.2 percentagepoints.

On Frequency-Wise Normalizations for Better Recording Device Generalization in Audio Spectrogram Transformers