@atriantafy.bsky.social
Found it pretty good. Quality felt Interspeech-y. Not sure if that's bad for IS or good for ITG. But I still got the feeling we're all playing variations of the same tune
Got into ITG (www.tu.berlin/qu/itg-sc-2025) with "Speaker vs Noise Conditioning for Adaptive Speech Enhancement". Built on DFingerNet (arxiv.org/abs/2501.10525) by comparing speaker vs noise adaptation. Nobody had done that before 1-on-1. Results mixed. We'll keep pushing this..
ITG Conference on Speech Communication 2025 - TU Berlin
www.tu.berlin
September 26, 2025 at 7:50 PM
Found it pretty good. Quality felt Interspeech-y. Not sure if that's bad for IS or good for ITG. But I still got the feeling we're all playing variations of the same tune
arxiv.org/abs/2508.02448
One of the key parts of my dissertation is now out as a preprint: arxiv.org/abs/2508.02448 (& under review)
How much progress have we made on speech emotion recognition using deep learning? The answer is not so easy to measure...
#SER #Audio #AI #DeepLearning
One of the key parts of my dissertation is now out as a preprint: arxiv.org/abs/2508.02448 (& under review)
How much progress have we made on speech emotion recognition using deep learning? The answer is not so easy to measure...
#SER #Audio #AI #DeepLearning
Charting 15 years of progress in deep learning for speech emotion recognition: A replication study
Speech emotion recognition (SER) has long benefited from the adoption of deep learning methodologies. Deeper models -- with more layers and more trainable parameters -- are generally perceived as bein...
arxiv.org
September 22, 2025 at 12:50 PM
arxiv.org/abs/2508.02448
One of the key parts of my dissertation is now out as a preprint: arxiv.org/abs/2508.02448 (& under review)
How much progress have we made on speech emotion recognition using deep learning? The answer is not so easy to measure...
#SER #Audio #AI #DeepLearning
One of the key parts of my dissertation is now out as a preprint: arxiv.org/abs/2508.02448 (& under review)
How much progress have we made on speech emotion recognition using deep learning? The answer is not so easy to measure...
#SER #Audio #AI #DeepLearning
Made it back to P-IEEE!(proceedingsoftheieee.ieee.org/april-2025/)
A. Triantafyllopoulos et. al, "Computer Audition: From Task-Specific Machine Learning to Foundation Models," in Proceedings of the IEEE, vol. 113, no. 4, pp. 317-343.
It's an overview of FMs for #audio.
#AI
A. Triantafyllopoulos et. al, "Computer Audition: From Task-Specific Machine Learning to Foundation Models," in Proceedings of the IEEE, vol. 113, no. 4, pp. 317-343.
It's an overview of FMs for #audio.
#AI
April 2025 - Proceedings of the IEEE
Regular Papers Volume 113, Issue 4 April 2025 Access the Full Issue Regular Papers Scanning the Issue Computer Audition: From Task-Specific…
proceedingsoftheieee.ieee.org
September 16, 2025 at 12:36 PM
Made it back to P-IEEE!(proceedingsoftheieee.ieee.org/april-2025/)
A. Triantafyllopoulos et. al, "Computer Audition: From Task-Specific Machine Learning to Foundation Models," in Proceedings of the IEEE, vol. 113, no. 4, pp. 317-343.
It's an overview of FMs for #audio.
#AI
A. Triantafyllopoulos et. al, "Computer Audition: From Task-Specific Machine Learning to Foundation Models," in Proceedings of the IEEE, vol. 113, no. 4, pp. 317-343.
It's an overview of FMs for #audio.
#AI
"AAAI-26 received almost 29,000 submissions"
And that's a good thing? We've broken peer review a long time ago. Can we please stop doing this to ourselves? How can we possibly build on previous research, if no human being can ever read it in a lifetime? #AAAI
And that's a good thing? We've broken peer review a long time ago. Can we please stop doing this to ourselves? How can we possibly build on previous research, if no human being can ever read it in a lifetime? #AAAI
August 29, 2025 at 11:50 AM
"AAAI-26 received almost 29,000 submissions"
And that's a good thing? We've broken peer review a long time ago. Can we please stop doing this to ourselves? How can we possibly build on previous research, if no human being can ever read it in a lifetime? #AAAI
And that's a good thing? We've broken peer review a long time ago. Can we please stop doing this to ourselves? How can we possibly build on previous research, if no human being can ever read it in a lifetime? #AAAI
After 2+ years of working on ecoacoustic for the biodiversity exploratories (www.biodiversity-exploratories.de/en/), we finally have our first publication out: www.sciencedirect.com/science/arti...
Soundscape-based evaluation of small-scale forest management interventions
While forests worldwide decline and European forests bear extensive management impacts, biodiversity within these ecosystems faces increasing threats …
www.sciencedirect.com
August 27, 2025 at 12:38 PM
After 2+ years of working on ecoacoustic for the biodiversity exploratories (www.biodiversity-exploratories.de/en/), we finally have our first publication out: www.sciencedirect.com/science/arti...
Got into ITG (www.tu.berlin/qu/itg-sc-2025) with "Speaker vs Noise Conditioning for Adaptive Speech Enhancement". Built on DFingerNet (arxiv.org/abs/2501.10525) by comparing speaker vs noise adaptation. Nobody had done that before 1-on-1. Results mixed. We'll keep pushing this..
ITG Conference on Speech Communication 2025 - TU Berlin
www.tu.berlin
August 4, 2025 at 8:23 PM
Got into ITG (www.tu.berlin/qu/itg-sc-2025) with "Speaker vs Noise Conditioning for Adaptive Speech Enhancement". Built on DFingerNet (arxiv.org/abs/2501.10525) by comparing speaker vs noise adaptation. Nobody had done that before 1-on-1. Results mixed. We'll keep pushing this..