Titulo Estágio
Melody Detection in Polyphonic Audio
Áreas de especialidade
Sistemas Inteligentes
Local do Estágio
DEI-FCTUC
Enquadramento
Melody extraction from polyphonic audio is a research area of increasing interest in Music Information Retrieval (MIR). It has a wide range of applications in various fields, including music information retrieval (particularly in query-by-humming, where the user hums a tune to search a database of musical audio), automatic melody transcription, performance and expressiveness analysis, extraction of melodic descriptors for music content metadata, and plagiarism detection, to name but a few. This area has become increasingly relevant in recent years, as digital music archives are continuously expanding. The current state of affairs presents new challenges to music librarians and service providers regarding the organization of large-scale music databases and the development of meaningful methods of interaction and retrieval.
Objetivo
In [Paiva, 2006], the problem of melody detection in polyphonic audio was addressed following a multistage approach, inspired by principles from perceptual theory and musical practice. The system comprises three main modules: pitch detection, determination of musical notes (with precise temporal boundaries, pitches, and intensity levels), and identification of melodic notes.
The main objective of this thesis is to build on the work carried out in [Paiva, 2006] to tackle several open issues in the developed system, namely: derive a more efficient pitch detector, improve note determination in the presence of complex dynamics such as strong vibrato, address the current limitation in the melody/accompaniment discrimination task, improve the reliability of melody detection in signals with lower signal-to-noise-ratio, add top-down information flow to the system (e.g., the effect of memory and expectations), add context information (e.g., piece tonality, rhythmic information), augment the song evaluation database etc. Machine learning methodologies will be key to attain the objectives.
Plano de Trabalhos - Semestre 1
1. Literature review: music and melody, melody extraction methods and machine learning methodologies
2. Dataset acquisition and/or improvement of current one
3. Formulation of research questions
4. Formulation of hypotheses to answer the research questions
Plano de Trabalhos - Semestre 2
5. Development and evaluation of the proposed algorithms
6. Thesis writing
7. Possible writing of a scientific research paper
Condições
The MSc candidate will join an active research group, which will give him strong scientific support, in an excellent work atmosphere.
Moreover, the candidate will have the opportunity to work in a cutting-edge research area with several open and exciting research possibilities, with plenty of room for scientific innovation.
Payment: unavailable
Observações
References:
Rui Pedro Paiva, “Melody Detection in Polyphonic Audio”, PhD Thesis, Department of Informatics Engineering, University of Coimbra, 2006, Portugal.
(inserido pelo coordenador do MEI)
Não foi fornecida informação sobre se o estágio é ou não remunerado.
Orientador
Prof. Dr. Rui Pedro Paiva
ruipedro@dei.uc.pt 📩