Digital Circuits and Systems Group: Best Paper Award at the tinyML Research Symposium 2024
The Digital Circuits and Systems Group, lead by Prof. Luca Benini, received the Best Paper Award at the tinyML Research Symposium 2024. The paper, titled "Boosting Keyword Spotting through On-Device Learnable User Speech Characteristics", was co-authored by Cristian Cioflan and Lukas Cavigelli.
The paper shows that AI-based speech models can be augmented with simple user features, offering better recognition rates for different accents or voice pitches. Their method is suitable for tinyML devices, as it needs only 0.7 mJ of energy to improve the accuracy results by 78% of what training the entire model achieves.
Paper: external page Boosting keyword spotting through on-device learnable user speech characteristics
Congratulations!