Fraunhofer Center for Assistive Information and Communication Solutions – AICOSDescription:
The student will study different audio signal processing techniques, extract relevant features from speech recordings, and machine learning methods to explore the relationship between extracted features and specific outcomes. These methods can include unsupervised learning techniques to find groups of patients with similar progression rates or gradual dissimilarities through time, for instance, or supervised methods to predict a functional score related to the bulbar system, or study how early speech-related features can predict ALS diagnosis. Furthermore, the student will explore different types of representation learning to assess their potential to find representative feature spaces, which might bring new, interpretable insights into the disease.
Outcome:
This thesis is aligned with the HomeSenseALS project. The student will develop speech degeneration analysis methodologies, including patient stratification, diagnostic biomarkers, interpretable speech features (XAI), and gradual changes.
Author: Ricardo Cebola
Type: MSc thesis
Partner: FCT NOVA – Faculdade de Ciências e Tecnologia da Universidade Nova de Lisboa
Year: 2022