Early recognition of Parkinson's through speech technology

Parkinson’s disease is the fastest growing age-related neurological disease in the world. A monotonous voice is one of the typical characteristics that can indicate Parkinson’s. University of Groningen/Campus Fryslân PhD student Vass Verkhodanova investigated how to notice changes in speech and how this can lead to the earlier detection and treatment of Parkinson’s.

The change in speech is one of the first symptoms of Parkinson’s disease and appears earlier than other known symptoms such as stiffness or tremor. If you can detect these changes earlier, you can also start speech therapy at an earlier stage of the disease progression. However, diagnosing Parkinson’s is not that easy: there is not yet a simple test one can do. Verkhodanova’s research shows that speech can play a key role in this. In her thesis she investigated how people can recognise speech of people with Parkinson’s. The ultimate goal is to use speech technology to detect Parkinson's disease at an early stage.

Verkhodanova conducted a large-scale study in which she played audio fragments of both neurologically healthy Dutch speakers and Dutch speakers with Parkinson's to listeners with no experience in speech disorders and to speech therapists. Her findings showed that the international group, namely the German and English listeners who did not speak Dutch, were more accurate at picking out the 'unhealthy' speech fragments. Their unfamiliarity with the language allowed them to ignore the content of the message and focus on the acoustics. In other words, they relied on “how” speech sounded rather than “what” has been said.

App for language training isolation

Currently, Verkhodanova is working with other scientists on developing an app that would allow people with Parkinson’s to train their speech as it deteriorates. The app differs from other existing apps in that it is also suitable for speakers of minority languages. Starting speech therapy early can help prevent people with Parkinson’s from becoming socially isolated.

About Vass Verkhodanova

Verkhodanova has a B.A. in Computational and Applied Linguistics and a M.A. in Phonetics and Speech Technologies from the Saint Petersburg State University, Russia (both cum laude). For her Master’s thesis, Verkhondanova worked on an algorithm for automatic detection of hesitation phenomena in Russian Spontaneous speech. During her Master’s, she was hired as a researcher in Speech and Multimodal Interfaces laboratory at Saint Petersburg Institute for Informatics and Automation of the Russian Academy of Sciences and continued working on the topic of hesitations in Russian spontaneous speech. During her time there, Verkhodanova received a patent for her algorithm and won several prestigious grants for her research, including the Young Research Grant from Russian Foundation for Basic Research (received twice), the Young Researcher Grant for Basic Research (received three times). Additionally, she worked and published on various projects dealing with speech technologies and phonetics of minority and under-resourced languages. For the last two years, she also has been a programme committee member for the SPECOM conference. Verkhodanova also co-supervised multiple interns and master students on topics related to voice technology.