Joseph Fourier Prize awarded to Kateřina Žmolíková for her work in the area of speech separation

Kateřina Žmolíková, a doctoral student at the Faculty of Information Technology of BUT, has been awarded the prestigious Joseph Fourier Prize today. She ranked third for her work in the area of speech separation. The prize, which is awarded to doctoral students for successful work in the area of computer science and information technology, was presented to her at the French Institute in Prague by French ambassador Roland Galharague.

In her work, Kateřina focuses on extraction of a single speaker from a speech mixture. "We are trying to extract speech of a single specific speaker from a recording in which multiple people speak at once and remove the other speakers. This can be used as a pre-processing step for various other technologies, such as speech-to-text conversion the accuracy of which is otherwise markedly decreased if there are multiple speakers speaking in a recording," Kateřina explains and further adds that the work was done largely in co-operation with the NTT research laboratory in Japan.

The technology Kateřina is working on is widely used today. Digital personal assistants like Alexa or Google Home are very popular nowadays, but the technology can also be used in applications for people with hearing impairment or for automated subtitling.

The Joseph Fourier Prize is an annual competition organised by the French Embassy in Prague in co-operation with Atos. Each year, authors of the best works go head-to-head in the competition focusing especially on design and use of compute-intensive algorithms and methods, simulations and modelling and on processing of large data volumes. More than a dozen of contestants from the top Czech universities participated in this year's competition. The winners receive financial support and are presented with the opportunity to travel to a research laboratory in France for scholarship. Last year, Vojtěch Mrázek was awarded the first place for his work focusing on the use of machine learning for optimisation and approximation of numerical circuits.