Unit:
Department of Informatics and TelecommunicationsΠληροφορική
Author:
ASLANIDIS THEOFANIS
Supervisors info:
Παναγιώτης Σταματόπουλος, Επίκουρος καθηγητής, Τμήμα Πληροφορικής και Τηλεπικοινωνιών, ΕΚΠΑ
Original Title:
Deep Learning in Audio Chord Estimation
Translated title:
Deep Learning in Audio Chord Estimation
Summary:
Each music piece consists of a set of different audio chords. These chords are the song’s
foundation and a skilful musician can identify them by ear. Although, most musicians can
identify audio chords, most non musically trained people cannot recognize them. This
thesis researches the use of neural networks and their importance, in the process of
identifying audio chords. Neural networks have shown great application on identifying
objects, as well as on extracting contextual information through time. The combination of
those characteristics is what this thesis will explore. More specifically, in this thesis what
is going to be presented is the power of a recurrent convolutional neural network in
comparison to other architectures for the purpose of identifying objects (chords) that have
an association through time.
Main subject category:
Technology - Computer science
Keywords:
Neural networks, Deep Learning, Convolutional Neural Networks, Recurrent Neural Networks, R-CNN, Audio Chord Estimation
