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/

Zendy Plus

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

Zendy Tools

Zendy Open

When convolutional neural networks are used to tackle learning problems basedon music or, more generally, time series data, raw one-dimensional data arecommonly pre-processed to obtain spectrogram or mel-spectrogram coefficients,which are then used as input to the actual neural network. In thiscontribution, we investigate, both theoretically and experimentally, theinfluence of this pre-processing step on the network's performance and pose thequestion, whether replacing it by applying adaptive or learned filters directlyto the raw data, can improve learning success. The theoretical results showthat approximately reproducing mel-spectrogram coefficients by applyingadaptive filters and subsequent time-averaging is in principle possible. Wealso conducted extensive experimental work on the task of singing voicedetection in music. The results of these experiments show that forclassification based on Convolutional Neural Networks the features obtainedfrom adaptive filter banks followed by time-averaging perform better than thecanonical Fourier-transform-based mel-spectrogram coefficients. Alternativeadaptive approaches with center frequencies or time-averaging lengths learnedfrom training data perform equally well.

Basic Filters for Convolutional Neural Networks Applied to Music: Training or Design?