C4DM Seminar: Pablo Alejandro Alvarado Duran - Physically-Musically Inspired Probabilistic Models for Audio Content Analysis

For external participants: Please join our mailing list to receive announcements of future C4DM seminars.

Date and Time
Wednesday, 10th February, at 3:00pm

Place
Room ENG209, Electronic Engineering building, Queen Mary University of London, Mile End Road, London E1 4NS. Information on how to access the school can be found at here.

Speaker
Pablo Alejandro Alvarado Duran

Title
Physically-Musically Inspired Probabilistic Models for Audio Content Analysis

Video

Abstract
The aim of audio content analysis in music information research is to estimate musical concepts such as pitch, melody, chords, onset, beat, tempo and rhythm, which are present but hidden in the audio signal. Automatic music transcription (AMT) refers to extraction of a human readable and interpretable description from a recording of a music performance.

Real music signals are highly variable, nevertheless they have strong statistical structure. Prior information available about the underlying structures, such as knowledge of physical mechanisms by which sounds are generated and rules by which complex sound structure is compiled (notes, chords, a complete musical score), can be naturally unified using Bayesian modelling techniques. Typically algorithms for AMT independently carry out individual tasks such as multiple-F0 detection, and beat tracking. The challenge remains to perform joint estimation of all parameters.

We present the design, initial implementation, and results of a Bayesian approach for audio content analysis. The proposed methodology based on Gaussian processes seeks joint estimation of multiple music concepts by incorporating in the kernel prior information about the mechanistic and non-stationary behaviour, rich spectral content, and musical structure, present in the modelled music signal.

Bio
Pablo A. Alvarado is interested in probabilistic approaches for modelling music signals, with a focus on Gaussian processes and kernel methods. Pablo holds an Electronic Engineering degree from Universidad Tecnológica de Pereira, Colombia (UTP), and a MSc in Electric Engineering from the UTP. Pablo is currently a member of C4DM pursuing a PhD at Queen Mary University of London with Dan Stowell.