Department of
Computational Perception
| |
Department of
Computational Perception
|
|
MUSIC SIMILARITY, ARTIST & INSTRUMENT DETECTION
Audio search engines, music broadcasting services and music composing softwares are examples of applications for audio-based music similarity. In music similarity estimation, similarity factors can be estimated in a high-dimensional and information-rich space based on the desired similarity factors such as rhythm, genre, singing voice or instrumentation. In our institute, we are developing similarity estimation methods to capture different attributes such as rhythm, timbral perspective etc using different machine learning approaches such as factor analysis.
Requirements: We are looking for passionate students interested in machine learning and music similarity to work with us on this subject. Matlab programming skills, machine learning background, a basic knowledge about signal processing is required. Also, having a knowledge in python programming and being familiar with music and the characteristics of different genres is appreciated. The students will be provided with related datasets and state-of-the-art implementations of music similarity algorithms and will be asked for further investigations about new techniques and applications.
Music artist recognition is the task where a music artist (e.g. singer) is recognized from a song. This task becomes more challenging when it comes to live performances where machine learning approaches are required to detect different bands playing live. We use state-of-the-art techniques in machine learning for music artist recognition in large scale and noisy environment and also we try to provide solutions for real-world applications such as smart-phone apps.
Requirements: Students interested in machine learning and music artist recognition can contact us for further information about working on this topic. Programming skills in Matlab, and having background in machine learning is required. Having a basic knowledge about signal processing and statistics can be very useful for students working on this project. The students will be provided with related datasets and state-of-the-art implementations of music artist recognition algorithms and will be asked about investigations for new techniques and applications.
Contact: Hamid Eghbal-zadeh, Bernhard LehnerWhen it comes to classical music, knowing when and where the instruments are active is always an interesting knowledge which can assist classical music fans by providing an on-line or off-line visualization such as instrument activity plot. Also, having the information about the active instruments in a music excerpt can is useful in other areas such as source separation, music transcription or score following. In cp, we provide solutions for musical instrument detection with a focus on classical music.
Requirements: For this task, we are looking for students that are interested in machine learning and musical instrument detection. Programming skills in Python and having background in machine learning is required. Also, being familiar with Neural Networks and Deep Learning and having a basic knowledge about signal processing is appreciated. The students will be provided with related datasets and state-of-the-art implementations of musical instrument detection and will be asked for investigations about new techniques and applications.
Contact: Hamid Eghbal-zadeh