Main aims are:
- to bring together in particular (but not solely) the nice bunch of people at RITMO working on topics more or less closely related to signal processing, especially in a multimodal context.
- to offer a forum where RITMO’s methodological needs in multimodal signal processing can be addressed. Indeed, RITMO features a large number of data-centric projects, where significant data need to be processed further.
- to get a good overview of, and articulate, ongoing works, to foster collaborations.
For anyone interested, please subscribe to the mailing list:
https://sympa.uio.no/ritmo.uio.no/info/msp-list
Current season:
Meeting at RITMO room v217, on some Tuesdays from 12:15 to 13:00. (Roughly biweekly, alternating with?Interaction and Robotics cluster meetings)
The main presentation starts at 12:15, but feel free to arrive earlier for a chat.
It is also possible to join via Zoom.
3.2. Classification vs. Active detection (Finn)
Classical machine learning and standard statistical methods identify reliable associations between variables in controlled dataset with clean(er) sampling conditions and non-causal features evaluated over epochs with consistent dimensions. Such results often inspire interest in active detection processes: systems that can evaluate related (though rarely identical) information in ongoing recording conditions with much wider variability. However adapting low resolution pattern detection in controlled condition to active sensing is often difficult and not always possible. And even in situations where the signal processing doesn't not need to be online, replicating patterns from tidy datasets in larger collections is not always smooth.
For discussion:
- What kind of signal processing are we conducting? What experience do we have with these different evaluation conditions?
- What are the main challenges in developing active sensing capacity from successful classification results?
- Are there signals or contexts in which it is not possible to develop active detection processes?
- Is translation from classification to active detection more of a hurdle for some practices than others?
- How should we talk about our research to make these distinct signal processing conditions more accessible to non-experts?
3.3. Ambisonics visualization toolbox (Arthur)
17.3. Integrating our toolboxes
- MiningSuite (including MIRtoolbox) (Olivier)
- Activity Analysis and Respy Libraries (Finn)
- MGT (Alexander+Balint)
- Pixasonics (Balint)
- Ambiviz (Arthur)
12.5. Multimodal quantity of motion (Hugh)
26.5. TBD
?
Previous seasons:
August 2025:
16.9. Opening meeting of the season
agenda?gdoc
(15.10. Workshop with Petri Toiviainen)
organised by Laura, cf. her email
28.10. Round table
25.11.?Video encoding
Hugh’s project on testing video encodings for their impact on video based motion evaluations.
Discussion: Parallel issues within audio or for other video features?
9.12. Sagar presenting Anyone interested in presenting? Please contact us.
Spring 2025:
4.2. Opening meeting of the season
(4.3.?RITMO Workshop on Music and AI)
1.4. Frequency, periodicity, and oscillating systems: modeling and analysis across time scales
Our favourite tools or techniques or salient issue with handling periodicity in our areas of research.
29.4. Sound synthesis
27.5. Video-based approaches
10.6. Datasets day
Fall 2024:
10.9: Introduction
24.9: DjembeDance data analysis (Sagar)
22.10:?Differentiable Digital Signal Processing (Riccardo Simionato, IMV)
6.11:?Still-standing data analysis (Arthur Jinyue)
19.11:?Bodies in Concert data analysis (Hugh)
4.12:?Music Technology workshop with Xavier Serra (RITMO SAB)
RITMO researchers briefly presenting works on topics related to Music Technology (audio signal processing, music information retrieval, musical interfaces, computational musicology, etc.), and time for Q&A and discussions. Program.
Spring 2024:
6.2: Introduction
5.3: DjembeDance project (Rainer and Sagar)
Slides:?PDF, PPTX (with video examples)