Important Announcements
Next meetup: Monday, September 18, 2017
Links
Events
2017-10-16: Deep Learning for Music Recommendation and Generation
Monday, October 16, 2017
7:15 PM, hosted by SoundCloud - Rheinsberger Str. 76/77 10115, Berlin map
Sander Dieleman: Deep Learning for Music Recommendation and Generation
The advent of deep learning has made it possible to extract high-level information from perceptual signals without having to specify manually and explicitly how to obtain it; instead, this can be learned from examples. This creates opportunities for automated content analysis of musical audio signals. In this talk, I will discuss how deep learning techniques can be used for audio-based music recommendation. I will also discuss my ongoing work on music generation in the raw waveform domain with WaveNet.
Sander Dieleman is a Research Scientist at DeepMind in London, UK, where he has worked on the development of AlphaGo and WaveNet. He was previously a PhD student at Ghent University, where he conducted research on feature learning and deep learning techniques for learning hierarchical representations of musical audio signals. During his PhD he also developed the Theano-based deep learning library Lasagne and won solo and team gold medals respectively in Kaggle’s “Galaxy Zoo” competition and the first National Data Science Bowl. In the summer of 2014, he interned at Spotify in New York, where he worked on implementing audio-based music recommendation using deep learning on an industrial scale.
2017-09-18: Phase Reconstruction from Magnitude Spectrograms
Monday, September 18, 2017
6:30 PM, hosted by Native Instruments - Schlesische Straße 29-30, 10997 Berlin map
Christian Dittmar: Phase Reconstruction from Magnitude Spectrograms
Phase reconstruction from magnitude spectrograms is an important pre-requisite for signal reconstruction in many audio processing applications. Among others, proper phase reconstruction can be beneficial for audio coding, speech enhancement, and source separation, i.e., whenever we process audio signals in the Short-time Fourier Transform domain. This talk covers the most relevant techniques, namely Phase-locked Vocoder, Phase Gradient Integration, Convex Optimization, and Iterative Phase Reconstruction, and discusses their advantages and drawbacks. Furthermore, I will point out some relationships between different methods and how relevant ideas spread across from distinct research directions. I will conclude this talk with some recommendations for the practical use of the investigated algorithms, especially focusing on music source separation.
Christian Dittmar received the Diploma degree in electrical engineering from the Jena University of Applied Sciences, Jena, Germany, in 2003. Since summer 2014, he has been working toward the Ph.D. degree in the research group of M. Müller, International Audio Laboratories Erlangen, Germany. Before that, he had worked at the Fraunhofer Institute for Digital Media Technology (IDMT), Ilmenau, Germany. The institute is headed by Prof. K. Brandenburg, one of the co-inventors of the MP3 audio compression format. From 2006 to 2014, Christian Dittmar was the Head of the Semantic Music Technology Research Group, IDMT. Since 2012, he has been also the CEO and co-founder of the music technology start-up Songquito. In 2014, he received the AES Citation Award for his involvement in organizing the 42th and 53th AES International Conference on Semantic Audio. In 2014, he was nominated as one of the 39 visionaries of the digital society in Germany by the Gesellschaft für Informatik. During the 2015 HAMR@ISMIR hackathon, he received the Best Hack Award. He authored and coauthored a number of peer-reviewed papers on music information retrieval topics. His recent research interests include music information retrieval, audio signal processing, and music education applications.
2017-07-17: Corpora for MIR Research & Computational Music Theory
Monday, July 17, 2017
7:30 PM, hosted by CRCLR - map
this meetup is supported by Aitokaiku
Xavier Serra: Creating and maintaining corpora for MIR research
One of the biggest bottlenecks for the advancement of the research in MIR is the lack of very large and openly available corpora of music data, corpora with which to train and test machine learning models. In this presentation, I want to talk about this issue and about the initiatives in which my group is involved to tackle it. I will cover freesound.org and the recent effort to create the FreesoundDataset. I will also describe our efforts in collaborating with MusicBrainz to create AcousticBrainz.org, a framework to crowdsource acoustic information for music tracks. Finally, I will talk about our research in CompMusic to develop Dunya, which comprises corpora of several music repertoires plus software tools, for the purpose of musicological research.
Xavier Serra is Associate Professor of the Department of Information and Communication Technologies and Director of the Music Technology Group at the Universitat Pompeu Fabra in Barcelona. After a multidisciplinary academic education, he obtained a PhD in Computer Music from Stanford University in 1989 with a dissertation on the spectral processing of musical sounds that is considered a key reference in the field. His research interests cover the computational analysis, description, and synthesis of sound and music signals, with a balance between basic and applied research and approaches from both scientific/technological and humanistic/artistic disciplines. Dr. Serra is very active in fields of Audio Signal Processing, Sound and Music Computing, Music Information Retrieval and Computational Musicology at the local and international levels, being involved in the editorial board of a number of journals and conferences and giving lectures on current and future challenges of these fields. He was awarded an Advanced Grant from the European Research Council to carry out the project CompMusic aimed at promoting multicultural approaches in music information research. More info: https://www.upf.edu/web/xavier-serra
Ryan Groves: Computational Music Theory: Working with Symbolic Data
This talk will present an overview of methods that look to extract information from symbolic representations (such as digital scores). The tasks discussed will include pattern recognition (e.g., melodic phrase detection), harmonic sequence analysis and melodic reduction.
Ryan received his B.S. in Computer Science from UCLA, and continued on to complete a Master’s in Music Technology from McGill University. As the former Director of R&D for Zya, he developed a musical messenger app that automatically sings your texts, called Ditty. Ditty won the Best Music App of 2015 by the Appy Awards. In 2016, his research in computational music theory was awarded the Best Paper of the most prominent music technology conference, ISMIR. With his new venture, Melodrive, he and his co-founding team of two PhDs in Music and AI are looking to build the world’s best artificially intelligent composer, and to change the way music is experienced in video games and virtual environments.
2017-06-26: Source separation and Sample Detection
Monday, June 26, 2017
6:30 PM, hosted by Ableton Schönhauser Allee 6-7, Berlin
For the fourth Berlin MIR meetup, we’re happy to announce the following speakers:
Gerald Schuller: Multi-Channel Source Separation using Independent Component Analysis
Gerald Schuller is a professor at Technical University Ilmenau, where he heads the Applied Media Systems group and is associated with the Fraunhofer Institute for Digital Media Technology (IDMT). His research focusses on audio and signal processing, in particular audio and speech coding, filter banks, music information retrieval, and source separation. Gerald will present recent works on multi-channel source separation using independent component analysis (ICA).
Alexander Lerch: Drum Transcription and Sample Detection with Non-Negative Matrix Factorization
Non-Negative Matrix Factorization (NMF) is a popular and comparably simple technique that has been successfully applied to MIR tasks such as pitch transcription and source separation. In this talk, we will discuss applying NMF to the task of transcribing drum events from polyphonic mixtures and to the task of detecting a snippet of a song (sample) in a new mix.
Alexander Lerch is an assistant professor at Georgia Tech University and co-founder of zplane development, Berlin. Previously he was a lecturer at the audio communications department of the Technical University Berlin. His research focusses on music information retrieval and digital signal processing.
2017-04-24: Harmonic Mixing & Rhythmic Similarity
Monday, April 24, 2017
7:30 PM, hosted by Audio Communication Group of TU Berlin
Marchstraße 8, 10587 Berlin, Berlin - lab
For the second edition of the Music Information Retrieval meetup, the Audio Communication Group of TU Berlin is hosting us in the HybridLab.
Roman Gebhardt: Psychoacoustic Approaches to Harmonic Mixing
Roman Gebhardt (TU Berlin) will give a talk on his works on harmonic analysis of music based on psychoacoustic principles. He will present a system for automatic harmonic adjustment of two or more pieces of music that builds upon the theory of roughness. As a counterpart to well-known beat synchronization, this motivates synchronization for music mixing in the spectral domain to maximize the musical consonance of a mix. He will as well give insight into a perceptually based filtering process for chroma-analysis of musical audio that has proven successful in key detection tasks and outline the idea of an extension of classic key detection to root notes.
Nico Lehrbach: Measuring Rhythmic Similarity of Music Excerpts
Nico Lehrbach (TU Berlin) will present his work investigating the rhythmic similarity estimation of short music excerpts. Rhythmic similarity is seen as one component of general music similarity. Derived from existing measures of music similarity like the Rhythmic Patterns and the Onset Patterns, he presents a new approach, the Bar Patterns. A survey was carried out in order to gain a quantitative, empirical ground truth of the rhythmic similarity of short music excerpts presented to the participants. These empirical data indicate the general possibility of measuring rhythmic similarity and is taken as a base to compare the algorithmic approaches.
2017-03-20: Kick-off with Meinard Müller
Monday, March 20, 2017
7:30 PM, hosted by SoundCloud
Rheinsberger Str. 76/77 10115, Berlin - Reception is on the second floor of the factory building
We are very pleased to have Meinard Müller from the International AudioLabs Erlangen as our first speaker for the MIR meetup. He will give an overview of the current state-of-the-art in Music Information Retrieval and an outlook on future directions and research problems. After the talk, there will be the opportunity for personal conversations with him and some members of his research group as well as representatives from various Berlin/Potsdam based university groups and companies working on MIR.
Speaker
Meinard Müller is professor for Semantic Audio Processing at the International Audio Laboratories Erlangen, Germany, a joint institution of the Friedrich-Alexander-Universität Erlangen-Nürnberg (FAU) and the Fraunhofer Institute for Integrated Circuits IIS. His research interests include music processing, music information retrieval, audio signal processing, content-based multimedia, and motion retrieval.
https://www.audiolabs-erlangen.de/fau/professor/mueller
Contact Us
This Meetup is organized by a few people in the Berlin area, from both industry and academia.
You can reach the group using the mailing list of our google group.