Greetings from Kos

In September, doctoral student Jonathan Svirsky traveled to Kos, Greece, to attend Interspeech 2024, the world's leading conference in the field of speech signal science and technology. "This is a top-tier annual conference that has been held for 25 years now. It focuses on advancing automatic speech recognition and analysis capabilities across various applications. This year, the conference chair was Prof. Sharon Gannot from Bar-Ilan University's Faculty of Engineering. “It was supposed to be held in Israel, but because of the war, it was moved to Greece," Swirsky explains.

Svirsky, 36, married +5, is a PhD student of Dr. Ofir Lindenbaum’s. Their joint research is concerned with developing a lightweight model for distinguishing words from audio. "The model is designed to run on edge devices like microcontrollers. We managed to reduce the model's size by applying sparsification techniques to the audio features–thinning out non-essential information to enable more efficient and accurate data processing in neural networks. The model marks the audio region where the word is identified and predicts its category."

The paper was originally accepted to be presented in front of an audience, but due to flight constraints, Svirsky presented a poster instead. "The paper garnered significant interest from companies working in the field. Additionally, I had the opportunity to meet several leading researchers in the field and learn about other work presented at the conference."