Understanding and Evaluating Speech-Rate Feedback for Home Based Dysarthria Training

Sammanfattning

Speech rate is a central factor influencing speech intelligibility in individuals with dysarthria. While speech-language therapy provides structured feedback in clinical environments, individuals practicing at home often lack objective and understandable guidance regarding their speaking rate. Existing speech analysis tools are either technically complex and designed for researchers or lack scientifically validated measurement methods. This creates a gap between validated acoustic measurement techniques and accessible user-centered feedback tools for home-based speech training. The purpose of this thesis is therefore to investigate how speech-rate feedback can be measured and presented in a simple and understandable way for home-based dysarthria training. The research question guiding the study is: How can speech-rate feedback be measured and presented in a simple and understandable way for home-based dysarthria training? To address this question, a prototype web application was developed using Java for recording and interface design, and Praat scripting for acoustic syllable nucleus detection based on the method proposed by de Jong and Wempe (2009). Speech rate is calculated in syllables per second using active speaking span. The prototype provides categorized feedback (“Too fast”, “Good pace”, “Too slow”) through a user-friendly graphical interface. The assessment of the prototype took place through two online surveys, which had structured questions. System Usability Scale (SUS) testing together with six application specific questions on speech-rate feedback comprehension, was performed by 17 general users and patients. The two speechlanguage therapists (n = 2) performed a survey, which measured clinical relevance through nine different assessment items. The general user evaluation produced a mean SUS score of 81.2, which showed “Good” usability (Bangor et al., 2009), because users rated both the recording process and their entire experience at the highest level. The clinical evaluation produced a clinical relevance score of 4.28 out of 5 because users rated both personalised target range configuration and practice reminders at their highest score of 5.00. Expert qualitative evaluation found design improvements through specific feedback display color differentiation and structured speaking task prompts and clearer post-recording navigation and audio playback functionality. The study shows that intensity-based syllable nucleus detection works as a technical solution, which enables people to measure their speech rate at home because the three-category visual feedback system is understandable to general users and clinically valuable to experienced speech-language pathologists. The OpenVox modular rehabilitation framework uses the prototype as a tested base, which supports its future development. The main contribution of this thesis is the design, implementation, and evaluation of a user-centered speech-rate feedback module for home-based dysarthria training within the OpenVox framework. The contribution is not the development of a new speech-rate measurement algorithm, but the integration of an existing acoustic measurement approach into an accessible rehabilitation prototype. The work contributes through the implementation of understandable visual feedback, therapist-supported configuration features, and an evaluation of usability and clinical relevance.