Best Student Paper Award for Acoustics Lab doctoral student Eloi Moliner
Aalto Acoustics Lab's doctoral student Eloi Moliner received the Best Student Paper Award at the IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP) on the Island of Rhodes, Greece, on Friday, 9 June 2023. ICASSP is an annual, week-long flagship conference of the Signal Processing Society of the IEEE, the largest international organisation of electrical and electronics engineers. This time, the ICASSP conference had more than 4000 participants.
Eloi has been a doctoral student at Aalto University School of Electrical Engineering since 2021 and is focusing in his research on deep learning methods for audio restoration. In this paper, Eloi collaborated with Prof. Jaakko Lehtinen of the Aalto Department of Computer Science and with Prof. Vesa Välimäki, who is Eloi’s supervisor.
The award-winning paper discusses how to solve inverse problems in audio processing. The proposed solution is based on a diffusion model, which uses a deep neural network for inference. This paper shows examples of correcting a clipped music recording, filling dropouts in a sound file, and enhancing the brightness of music by hallucinating the missing high frequencies. Diffusion models have become popular in AI research in the past few years and are famous for their miraculous results in image generation.
Congratulations to Eloi!
Read the article . Many audio examples and animations related to this work are available .
More information
Vesa Välimäki
Professor of audio signal processing
Department of Information and Communication Engineering / Acoustics Lab
School of Electrical Engineering
Aalto University
vesa.valimaki@aalto.fi
Read more news
Introducing Qi Chen: Trustworthy AI requires algorithms that can handle unexpected situations
AI developers must focus on safer and fairer AI methods, as the trust and equality of societies are at stake, says new ELLIS Institute Finland principal investigator Qi Chen
The research puzzle of when humans and AI don’t see eye to eye
Francesco Croce works on robustness in multi-modal foundation models
A rap algorithm led him to research language models at Google DeepMind – now Eric Malmi returns ÄûÃʵ¼º½ as an adjunct professor
Eric Malmi received his PhD from Aalto University in 2018 with a dissertation that developed AI methods for linking historical records and family trees. At Google DeepMind he has developed Gemini language models and a chess AI. He returned to his alma mater because of ELLIS Institute Finland.