Soner Türüdü will defend his PhD Thesis “Assessing Speech Recognition Abilities with Different Digits-in-Noise Implementations on Various Platforms and Populations”.
Supervisors: Prof. dr. ir. Deniz Başkent and Dr. Thomas Koelewijn
Abstract
Many people hear well in a quiet room but struggle to follow a conversation in a busy café or on a train platform. A quick way to measure this is the digits-in-noise test: listeners hear three spoken digits in background noise and type them in, while the noise becomes louder or softer depending on their answers. The test takes only a few minutes, can be done online and has been translated into many languages. But because every version is built a little differently, results from different studies are not always comparable. In his PhD research, Soner Türüdü investigated how the digits-in-noise test can be described, compared and interpreted more clearly. He developed an open description that lets researchers report exactly how their version works, so that others can rebuild it. Using this description, he showed that an online version gives almost the same results as testing in the lab, and that the way the sound is played to the two ears affects the score most. He also developed a Turkish version and found that the choice of speaker has a small but measurable effect. Finally, he showed that Turkish native speakers living in the Netherlands the test seems more difficult in Dutch than in Turkish, and that their Dutch proficiency explains this best. These findings help clinicians and researchers use the test with care, also for people tested in a language that is not their own.
