Current projects

  • Singing voice detection in musical signals

    More info soon!

  • Strategies for matching hummed queries

    Query-by-humming is a common topic in music information retrieval. In the query-by-humming task a hummed record, representing imprecisely a target melody, is given to an application which is supposed to retrieve the target melody from a dataset. Any algorithm addressing this task has to handle the deviations between query and target melodies in both time and frequency domains.

  • Audio Processing with Adaptive Computational Costs

    Audio processing applications can and do suffer from overload situations. This results in unwanted sound events: strong distortions, clicks and interruptions.

Previous projects

  • Reinforced Learning Strategies applied to Personalized Music Recommendation

    The number of songs available in digital format on the internet grows every day, and reaches an excessive volume. When deciding to listen to music, a user has such a large number of options at his disposal that he feels the need for a tool to help him in his decision-making process. Otherwise, it is plausible to believe that he would never access a large part of this material. Online platforms offer these users automatic recommendation services that analyze their listening history and keep available to them a sequence of songs that match their personal taste.

  • VORPAL - A middleware for real-time soundtracks in digital games

    Real-time soundtrack middleware project for digital games.

  • Speech processing to screen for phonological disorders

  • Musical brain-computer interfaces through EEG

  • Mobile Music and Musician

    Mobile music applications became increasingly common. Smartphones can now be used as digital instrument and provide musical creation with real-time control through interfaces, accessories and sensors. The advancement in telecommunications invited their use in musical collaboration and as musical controllers exchanging information between many communication channels. In this context, we can see that there are still many technologies from Computer Networks and Mobile Computing that can facilitate the process of communication between mobile devices in musical applications. Thus, this study aims at evaluating new technologies for the creation of collaborative musical environments.

  • Audio effects based on AM/FM decomposition

    This research explores the design of audio effects based on decompositions that take a music instrument signal and unravel it to the AM/FM representation. The AM/FM decomposition produces a pair of signals, also in the time domain, that represent the analyzed signal’s envelope (AM portion) and instantaneous frequency (FM portion). This pair of signals act jointly, and when used to modulate an oscillator’s amplitude and frequency are able to reconstruct the original signal. However, by manipulating the AM and FM portions new possibilities arise for signal processing and implementation of musical effects.

  • Melody and accompaniment audio signal separation

    Recovering perceptually valid information from audio signals is an important challenge with many areas of application. Among them is the computational auditory scene analysis, which seeks to achieve human performance in the recovery of sound characteristics. To this end, tools and methods that are widely studied in music information retrieval are used. This work uses a system developed to extract melodic information from signals and, from its results, performs the separation of this information in different audio tracks. The results show that, although simple, this method has some relevance in relation to what is perceived as main melody. There are several limitations, including the type of sound source and the quality of the melodic information returned by the system used.

  • Computational techniques applied to musical consonance and dissonance

    This undergraduate research aims to explore the concepts of consonance and dissonance in music and develop a validation experiment of the theory presented in Tuning, Timbre, Spectrum, Scale by William A. Sethares.

  • Musical accompaniment

  • Ensemble - A framework for multi-agent music systems

  • Cross-lingual Voice Conversion

    Voice Conversion is defined as the process of acoustical features conversion that sentences uttered by a source speaker are transformed into sentences that sound as sentences uttered by the target speaker.

  • Case studies on real-time audio processing on high-availability, low-cost computing platforms

    presentation-ajb.pdf defense-ajb.pdf thesis-ajb.pdf qualification-ajb.pdf

  • MOBILE - Interactive Musical Processes

    More infos at:

  • Medusa - A distributed audio environment

    Medusa is an audio/MIDI communication tool for local networks. The main goal is to unleash audio/MIDI communication between computers and software applications on a local area network without complex configurations or difficult set-ups.

  • JackTrip - Distributed music performances for home users

  • Short-term memory and working memory of melodic sequences

  • ACMUS: Room acoustics, simulation and design

    The AcMus project was started under the previous Thematic Project3 and had the AcMus program as one of its results. It is an application intended for the study, measurement and design of environments based on their acoustic properties. Two main modules are implemented: the Acoustic Measurement module, which extracts the Impulse Response of a room to later calculate a series of acoustic parameters of the same; and the Tools module, which offers various utilities such as room resonance mode calculator, reverberation time calculator, Schroeder diffuser design and convolution.

  • Musical structural segmentation methods based on sound descriptors

    Project description In this project, a comparative study of different music structural segmentation methods is presented, where the goal is to delimit the borders of musical sections and label them, i.e. group the sections that correspond to the same musical part. Novel proposals for unsupervised segmentation are presented, including methods for real-time segmentation, achieving expressive results, with error ratio less then 12%. Our method consists of a study of sound descriptors, an exposition of the computational techniques for structural segmentation and the description of the evaluation methods utilized, which penalize both incorrect boundary detection and incorrect number of labels. The performance of each technique is calculated using different sound descriptor sets and the results are presented and analysed both from quantitative and qualitative points-of-view.

  • Binaural auralization of moving sound sources using HRTFs

  • Audible Images: Creating Visual Outputs from Audio Inputs

  • ASyMuT: An automatic system for music transcription

    ASyMuT is a system for automatic music transcription (ie WAV to MIDI) and analysis. More information at:

  • Real-time distributed audio processing

    This project discusses and implements, under Linux, a mechanism for synchronous, distributed and low-latency audio processing on local area networks, allowing for the parallel processing of audio with relatively low-cost equipment. The primary objective is to make possible the use of distributed computing systems for music recording and editing in home or small-scale studios, but it is expected that the same mechanism is expandable to other media types (such as video) and to fields not related to multimedia that depend on distributed, synchronous and low-latency processing.

  • Andante: A Mobile Infrastructure for Music Agents

    Andante combines novel research in art and technology in search of new forms of artistic expression. Utilizing state of the art technology for mobile autonomous agents and musical synthesis, musicians, computer scientists, and amateurs experience a novel art form: musical agents.