A5088496136- Music and Audio Processing
- Music Technology and Sound Studies
- Speech and Audio Processing
- Speech Recognition and Synthesis
- Diverse Musicological Studies
- Advanced Adaptive Filtering Techniques
- Blind Source Separation Techniques
- Human Motion and Animation
- Natural Language Processing Techniques
- Hand Gesture Recognition Systems
- Video Analysis and Summarization
- Neuroscience and Music Perception
- Time Series Analysis and Forecasting
- Artificial Intelligence in Games
- Human Pose and Action Recognition
University of Padua
2008-2014
Sciences et Technologies de la Musique et du Son
2011
This article presents a gesture recognition/adaptation system for human--computer interaction applications that goes beyond activity classification and that, as complement to labeling, characterizes the movement execution. We describe template-based recognition method simultaneously aligns input templates using Sequential Monte Carlo inference technique. Contrary standard methods based on dynamic programming, such Dynamic Time Warping, algorithm has an adaptation process tracks variation in...
We present a methodology for the real time alignment of music signals using sequential Monte Carlo inference techniques. The problem is formulated as state tracking dynamical system, and differs from traditional Hidden Markov Model Dynamic Time Warping based systems in that hidden continuous rather than discrete. major contribution this paper addressing both problems audio-to-score audio-to-audio within same framework setting. Performances proposed on are then evaluated discussed.
The behavior of users music streaming services is investigated from the point view temporal dimension individual songs; specifically, main object analysis in time within a song at which stop listening and start another ("skip"). contribution this study ascertainment correlation between distribution skipping events musical structure songs. It also shown that such not only specific to songs, but independent cohort and, under stationary conditions, date observation. Finally, user behavioral...
A comprehensive methodology for automatic music identification is presented. The main application of the proposed approach to provide tools enrich and validate the descriptors recordings digitized by a sound archive institution. Experimentation has been carried out on three different datasets, including collection vinyl discs, although not linked particular recording carrier. Automatic allows digital library retrieve metadata about works even if information was incomplete or missing...
We present a system for automatic real time alignment of an acoustic music performance with digital representation its score, problem which is usually defined score following. The based on application hidden Markov models. A model automatically built from while decoding used to compute the most probable location along model. effectiveness proposed approach has been tested collection recordings orchestral music. Even if typical following accompaniment, in this paper we propose set novel...
This paper describes the implementation of a content-based cover song identification system which has been released under an open source license. The is centered around Apache Lucene text search engine library, and proves how classic techniques derived from textual Information Retrieval, in particular bag-of-words paradigm, can successfully be adapted to music identification. focuses on extensive experimentation most influential parameters, order find optimal tradeoff between retrieval...
We present FALCON, an open-source engine for content-based cover song identification written in Java. The popular Lucene search library is used as the core of software, proving that textual methods information retrieval can be successfully adapted to multimedia tasks. An overview system methodology and implementation are provided, along with experimental results on a medium-size test collection
Music source separation in the time-frequency domain is commonly achieved by applying a soft or binary mask to magnitude component of (complex) spectrograms. The phase usually not estimated, but instead copied from mixture and applied magnitudes estimated isolated sources. While this method has several practical advantages, it imposes an upper bound on performance system, where sources inherently exhibit audible "phase artifacts". In paper we address these shortcomings directly estimating...