Knowledge transfer from neural networks for speech music classification. - In: Music in the AI era, (2023), S. 202-213
A frequent problem when dealing with audio classification tasks is the scarcity of suitable training data. This work investigates ways of mitigating this problem by applying transfer learning techniques to neural network architectures for several classification tasks from the field of Music Information Retrieval (MIR). First, three state-of-the-art architectures are trained and evaluated with several datasets for the task of speech/music classification. Second, feature representations or embeddings are extracted from the trained networks to classify new tasks with unseen data. The effect of pre-training with respect to the similarity of the source and target tasks are investigated in the context of transfer learning, as well as different fine-tuning strategies.
Analyzing the potential of pre-trained embeddings for audio classification tasks. - In: 28th European Signal Processing Conference (EUSIPCO 2020), (2020), S. 790-794
In the context of deep learning, the availability of large amounts of training data can play a critical role in a models performance. Recently, several models for audio classification have been pre-trained in a supervised or self-supervised fashion on large datasets to learn complex feature representations, socalled embeddings. These embeddings can then be extracted from smaller datasets and used to train subsequent classifiers. In the field of audio event detection (AED) for example, classifiers using these features have achieved high accuracy without the need of additional domain knowledge. This paper evaluates three state-of-the-art embeddings on six audio classification tasks from the fields of music information retrieval and industrial sound analysis. The embeddings are systematically evaluated by analyzing the influence on classification accuracy of classifier architecture, fusion methods for file-wise predictions, amount of training data, and initial training domain of the embeddings. To better understand the impact of the pre-training step, results are also compared with those acquired with models trained from scratch. On average, the OpenL3 embeddings performed best with a linear SVM classifier. For a reduced amount of training examples, OpenL3 outperforms the initial baseline.
https://doi.org/10.23919/Eusipco47968.2020.9287743
Parameter extraction for bass guitar sound models including playing styles. - In: 2015 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), ISBN 978-1-4673-6998-5, (2015), S. 404-408
https://doi.org/10.1109/ICASSP.2015.7178000
Entwicklung eines parametrischen Instrumentencoders basierend auf Analyse und Re-Synthese von Gitarrenaufnahmen. - ca 130 S. : Ilmenau, Techn. Univ., Diplomarbeit, 2013
In dieser Diplomarbeit wird ein Algorithmus vorgestellt, der eine parametrische Audiocodierung von monotimbralen Gitarrensignalen ermöglicht. Als Grundlage dient dabei einzig das Ausgangssignal einer handelsüblichen Elektrogitarre. Anhand des digitalisierten Audiosignals werden Parameter extrahiert, die zum einen die automatische Notation des gespielten Stückes realisieren und zum anderen eine Synthese des notierten Stückes mittels eines Physical Modeling Verfahrens ermöglichen. Betrachtet werden sowohl monophone als auch polyphone Stücke sowie gängige Spieltechniken auf einer Gitarre. Weitere Bestandteile der Arbeit sind die Aufnahme und Annotation eines für diese Arbeit benötigten Datensatzes aus 261 Audiofiles sowie die anschließende Evaluation des Codierungsalgorithmus mittels eines einfachen MUSHRA Hörtestverfahrens zur Beurteilung der generierten Ergebnisse. Außerdem wird eine Lösung für ein mögliches Datenformat zur speicherschonenden Notation und Archivierung der Parameter vorgestellt.