Author: Lukas Weber, Bsc.
Title: Analyse, Design und Prototypische Umsetzung eines Serious Game für therapeutische Zwecke in der Sprachtherapie.
English title: Analysis, Design and Prototypical Implementation of a Serious Game to Aid in Speech Rehabilitation
Year of submission: 2025
Introduction
The topic of this thesis is how a serious game can support speech-language pathologists and patients in speech therapy. The game is designed to reduce the barriers to doing exercises at home and to increase the frequency with which patients practice these often monotonous exercises by making them more engaging. The author identifies two main issues of serious games in speech therapy: only whole words are included in the game and only whole words can be recognized by the voice recognition component. The author states, that in speech therapy processes patients often start by learning sounds and syllables, which are not covered by the game, before moving on to whole words. The aim of the diploma thesis is to analyse how this issue can be solved and to develop a working prototype. Furthermore, it aims to determine whether the data produced by the game might be of use for the therapist when measuring the therapy’s success.
In the beginning of the thesis there is a description of the problem and the expected result. Then there is a chapter that provides theoretical background on serious games, speech impediments, therapeutic compliance, neural networks, voice recognition and User Centred Design. Next, there is an overview over the state of the art of speech rehabilitation games. The author then describes the methodological approach, before moving on to the results. After that, there is a discussion of the research questions and a conclusion and future work. The thesis ends with the list of figures, list of tables, bibliography and appendix.
Evaluation of the Academic Quality
Quality of design and execution
The author applied an iterative User Centred Design process and collaborates with a speech therapist in six iterative development circles to integrate therapeutic requirements into the functionality of the game. Average Sound Completion Time (AST) and Most Common Mistake (MCM) are being implemented as metrics to make a connection between the functionality and the therapist’s needs. These points speak in favour of high executive quality, but there is one issue that cannot be overlooked: Despite developing a customized recognition model, its accuracy and processing speed remain insufficient in the end. The therapist explicitly states that the game cannot yet be used in actual therapy because of these shortcomings. This is a significant weakness because voice recognition is the game’s most essential component.
Considering the visual design of the workpiece, what first came to my attention is that the design seems a bit unfinished and unpolished, which makes it not very visually appealing. One can tell that the author of the thesis does not have a design background, which is why I think it would have been a good idea to learn the basic design principles. Having in mind that the author wrote this thesis in the field of Medical Informatics, I would not consider the shortcomings in visual design a major quality issue.
Another aspect that relates to the fact that this is not an interaction design thesis is that the interview questions in the appendix are rather shallow and do not get to the heart of the problem when it comes to the user experience and usability. One example for this is the first main question to the therapist (p. 137): “Was hältst du von dem Spiel?” (What do you think of the game?). The question is very general and does not address any specific features. Other examples for rather shallow questions are “Welche Verbesserungen würdest du empfehlen?“ (p .117) and „Wie hat dir der Entwicklungsprozess gefallen?“ (p. 118). The questions in the questionaire for the user testing are equally shallow, e.g. “Wie gut hat dir das Spiel gefallen?” (How much did you like the game?). I think the author could have benefitted a lot more of the interviews, had he asked more precise questions.
Degree of innovation
The thesis addresses a relevant gap in existing speech rehabilitation games. Its main contribution is the combination of three therapy stages (phonetic, syllabic and word-based) within one game, alongside an attempt to develop a voice recognition component specifically adapted to speech rehabilitation. The comparison of six existing Serious Games in Chapter 3 provides a foundation for identifying these shortcomings. The resulting solution also introduces additional features, particularly the therapist’s ability to create individualized categories and customize therapy levels. This offers greater flexibility than the existing solutions examined in the thesis. While making a contribution to the specific field of speech rehabilitation games in speech therapy, its innovation lies mainly in combining existing approaches with therapy-specific requirements and individual customization. However, the technical limitations of e.g. the voice recognition feature prevent some of its most innovative aspects from being fully realized.
Independence
The author identifies the research problem, establishes three research questions, conducts the literature review, develops the prototype and organizes the evaluation, which states a very independent working style. I would specifically like to point out the author’s initiative in recruiting participants to collect audio recordings for training the voice recognition model. This demonstrates an understanding of the technical challenges and a willingness to develop an individual solution rather than relying exclusively on existing software. The author also independently identifies and addresses technical problems during implementation, for example the limitations of the Whisper framework and difficulties exporting game data. At the same time the collaboration with the speech therapist has a major impact on design decisions, features and requirements. This does make sense when working with the approach of UCD, but I think the author would have benefitted from interviewing more therapists in the conceptual and design phase of the project.
Organisation and structure
The thesis follows a clear and logical structure, consisting of seven chapters that guide the reader from the initial problem definition to the final evaluation. I would like to point out the detailed documentation of the development process as a strength of this thesis. The six iterations are presented chronologically, making it possible to understand how feedback influenced the development of individual features.
However, the organisation could be improved in certain areas. Chapter 2 covers a broad range of topics, including speech disorders, neural networks and voice recognition. While these topics are relevant, some explanations are quite extensive and could be more directly connected to the actual design decisions. There is also some repetition between the introduction, theoretical background, results and discussion. For example, the limitations of existing speech recognition systems and the importance of covering multiple therapy stages are explained several times.
Quality of communication
The thesis communicates its central research objectives and technical development process in a generally understandable way. Complex technical concepts are supported by diagrams, tables, screenshots and practical examples. For example, the game architecture is explained through an overview of the gameplay process and a detailed description of the individual Unity scripts. The screenshots of the different interfaces also help readers understand the implemented features without having to interact with the prototype themselves. The author communicates limitations of his work quite clearly. Rather than presenting the prototype as a finished solution, he discusses problems with voice recognition, data collection and evaluation. This makes the work very credible and authentic. Nevertheless, some sections contain lengthy technical explanations that could be shortened, as for example the explanations of neural networks, that sometimes go beyond what is necessary to understand their application in the project.
The English writing style occasionally feels a bit informal or imprecise. Expressions such as “one can imagine” (p. 22, l. 7) and subjective formulations such as “the result can be considered acceptable” (p. 84 l. 1) could be replaced by more objective academic language.
Scope of the thesis
The thesis covers an ambitious range of topics and combines several disciplines, including speech therapy, artificial intelligence, game development, user-centered design and statistical evaluation.
The three research questions address different but interconnected aspects of the project:
- Identifying requirements for a Serious Game suitable for different stages of speech rehabilitation.
- Investigating which data can help therapists measure patient progress.
- Evaluating the subjective benefits of the developed game.
This broad scope is a strength because it considers the project from both technical and therapeutic perspectives. The author does not limit the work to developing a prototype but also investigates its potential application and evaluation.
However, the scope is arguably too ambitious for the available resources and timeframe. Developing a customized voice recognition model, implementing a complete game, conducting iterative design sessions and evaluating the prototype represent substantial individual tasks. This becomes visible in the quality of the user testings. Although 19 participants tested the game for at least one month, only one was an actual patient. Most participants had no direct experience with speech therapy.
Spelling, care and accuracy
The thesis demonstrates a generally high level of care in documenting the development process. The author provides detailed descriptions of the requirements, implementation decisions, technical problems and evaluation results. Tables, figures and appendices further contribute to the completeness of the documentation. The inclusion of the interview guide, questionnaire and questionnaire results is particularly positive because it allows readers to understand the research process and examine the evaluation material.
To assess the spelling of the paper, I prompted ChatGPT to search for spelling and language mistakes, so I did not have to examine the whole thesis in high detail. These are some examples of what came up:
- “vorangekokmmen” in the German acknowledgements (a typographical error).
- “in order to develop an later on evaluate” in Chapter 2 (grammatically incorrect).
- “a fracture of the therapy process” in the abstract, where “fraction” would be the appropriate word.
- Inconsistent capitalization and punctuation, as well as occasional awkward sentence constructions throughout the English text.
Use of literature
The thesis draws on a broad range of literature covering speech disorders, speech rehabilitation, serious games, voice recognition and user-centred design. This interdisciplinary approach provides a solid theoretical foundation. However, some sections mainly summarise existing research rather than critically comparing different perspectives. Overall, the literature is relevant and extensive, but a more critical synthesis would strengthen the academic foundation of the thesis.
Evaluation of the Practical Project / Work Produced
When looking through the paper, I could not find a link to the final prototype, so I must refer to the screenshots provided in the thesis. The practical implementation is connected very closely to the research questions. Especially the first research question (What are the requirements towards a Serious Game for speech rehabilitation that can be used across different steps of the rehabilitation process?) is strongly linked to the final prototype, since the research on the requirements had a big impact on the implemented features. The second research question (What logging data within a Serious Game can be used to measure player progress and provide meaningful feedback for speech therapists?) was answered consequently, but since most of the identified metrics rely on the voice recognition system, the contribution to the final prototype is questionable. The third research question (Does the game provide a subjective benefit within the rehabilitation process for a predefined test group?) is also closely connected to the workpiece, since it refers directly to the benefits of the game.
In some sections the link between the theoretical framework and the practical outcome could be stronger, for example, the technical background on neural networks (chapter 2) and voice recognition could be more closely connected to the specific design and implementation decisions.
Regarding the documentation of the workpiece, there are enough screenshots of the game to understand the concept. I as an UX-Design student would have wished for a more detailed user flow / user journey. Some of the technical sections are quite long, but since this is not an interaction design thesis, but a informatics thesis, I think the length and scope of those sections is appropriate.
In my opinion, the practical project meets the requirements of a Master’s thesis in terms of its concept, technical complexity and development effort. However, the implementation still has significant limitations, especially regarding the accuracy and speed of voice recognition. The evaluation is also too limited to demonstrate actual therapeutic effectiveness. Therefore, I would consider it a solid and ambitious prototype that meets the expectations of a Master’s thesis, but not a finished product ready for use in speech therapy.
Final conclusion
The thesis addresses a relevant and challenging topic by exploring how serious games can support speech rehabilitation. It combines theoretical research, technical development and user-centred design in an ambitious interdisciplinary project.
Strengths: One of the main strengths is the practical implementation. The author demonstrates considerable independence and technical knowledge in developing a game that covers different stages of speech rehabilitation. The iterative development process and regular collaboration with a speech therapist help ensure that the design is informed by practical requirements. The thesis is also clearly structured and documents the development process in considerable detail. Its literature review provides a broad theoretical foundation across several relevant fields.
Weaknesses: The main weakness is the gap between the project’s intended functionality and its actual technical performance. Voice recognition, which is central to the gameplay, does not yet work reliably enough for practical therapeutic use. Furthermore, the evaluation is limited in scope, particularly because only a small number of participants had relevant experience with speech therapy. Consequently, the thesis cannot convincingly demonstrate that the game improves therapeutic outcomes. Some theoretical sections could also be more critically connected to the practical implementation, and the academic writing contains occasional language and stylistic weaknesses.
Overall evaluation: Overall, I consider this a good and ambitious Master’s thesis with a strong practical and technical foundation. The author demonstrates independence, problem-solving skills and the ability to develop a complex interdisciplinary project. However, the limited reliability of the core voice recognition functionality and the lack of a more comprehensive evaluation weaken the project’s practical impact. It is therefore more convincing as a well-developed prototype and proof of concept than as a ready-to-use therapeutic tool. Despite these limitations, the thesis fulfils the expectations of a Master’s thesis through its scope, development effort and systematic approach.
Relevance to my own final thesis
One takeaway for my thesis was the structure and the theoretic framework. While I think that some of the contents of the theoretic framework may also be useful for my thesis, I think I would personally choose a different focus, which also may be related to the fact that I’m in a different study programme than the author. Instead of focusing on the specific speech impediments, I would in addition elaborate more on how the patients are impacted by the impediments and how this effects their needs when it comes to the user interface. It was interesting to see the interview questions in the appendix, but I would select more precise and in-depth questions to find out the core of the usability issues. In general I would include more research on the target group, like for example determine their pain points and gains as well as jobs to be done through interviews, define a user persona and rely on that information during the conceptual phase. I would also select test users that fit the target group better and consult more experts. Furthermore, I would do more desk research on the special requirements for applications in the health / therapy sector, not only from the therapists point of view, but also from an UX-theory point of view. Another thing that I liked about the thesis was how honest the author was with limitations and shortcomings. I think this is a better approach than trying to conceal possible shortcomings that might occur due to lack of time.
