Paul-Peter Arslan has completed a PhD combining data science, neuroscience and rehabilitation, supervised by Xia Xia, professor-researcher at ESILV, and co-supervised by Pavel Lindberg at the Institut de Psychiatrie et Neuroscience de Paris (IPNP). His research explores digital and AI-based methods for assessing and supporting the rehabilitation of manual dexterity after stroke.
From artificial intelligence and electromyography to music and augmented reality, the thesis brought several scientific and technological fields together around motor assessment and rehabilitation.
A PhD combining data science and neurorehabilitation
Paul-Peter Arslan defended his thesis, “Precision-Based Assessment and Rehabilitation of Manual Dexterity After Stroke Using Auditory Motor Paradigms,” following three and a half years of research.
Xiao Xiao, IFT Director and professor-researcher at ESILV, and Pavel Lindberg from the Institut de Psychiatrie et Neuroscience de Paris supervised the work. The PhD was co-financed by the De Vinci Higher Education and IPNP.
The research focused on using data science and interactive technologies to study manual dexterity, particularly in people recovering from a stroke. Its multidisciplinary scope connected neuroscience, rehabilitation, artificial intelligence, human-computer interaction (HCI) and music.
Measuring and rehabilitating manual dexterity after stroke
Paul-Peter Arslan’s thesis, “Precision-Based Assessment and Rehabilitation of Manual Dexterity After Stroke Using Auditory Motor Paradigms,” examines new ways to measure manual dexterity and support motor rehabilitation after stroke. The research pays particular attention to musical interfaces: rhythmic tasks require precise finger movements and sensorimotor synchronisation, two dimensions directly relevant to dexterity.
The thesis brings together three complementary research projects. Rhythm Karaoke (RK) uses a music-based finger-tapping task to assess timing precision. Tests conducted with healthy participants showed that familiar melodies could improve synchronisation and reduce perceived difficulty. Among stroke participants, melodic content and speech cues also improved timing precision, with speech cues producing the largest reduction in error (16%).
A second study combined high-density electromyography (HD-EMG) and artificial intelligence to analyse muscle activity during finger movements. A Transformer neural network was trained to predict continuous finger forces from EMG signals. The research also identified the contribution of the dorsal interossei muscles to predicting individual finger tapping. Despite reduced dexterity among stroke participants, the organisation of muscle activation patterns remained largely preserved.
The third project, ReTouche, examined piano practice and motor learning using a Yamaha Disklavier and multimodal feedback. Participants reported that visual replay helped them identify errors and focus on the learning process. Their continued engagement was associated primarily with visible progress and a sense of agency rather than game mechanics.
Together, these three projects provide methods for precision-based assessment and music-centred rehabilitation of manual dexterity after stroke, while opening further research directions for their integration into a common rehabilitation approach.
AI and interactive technologies applied to manual dexterity
Several projects were developed throughout the PhD. One involved a musical game in which tapping rhythms control vocal synthesis. The system was designed to support manual dexterity rehabilitation through auditory-motor interaction.
Another part of the research combined high-density electromyography (EMG) with artificial intelligence. This approach examined differences in hand-muscle co-activation between healthy participants and people who had experienced a stroke.
Paul-Peter Arslan also worked on an augmented reality system for piano learning, exploring how interactive digital interfaces can assist movement and learning. This research was presented at the ACM CHI Conference on Human Factors in Computing Systems.
Research experience at the MIT Media Lab
His doctoral work also led to a four-month research stay at the MIT Media Lab as a Visiting Research Student.
At MIT, Paul-Peter Arslan worked with Professor Hiroshi Ishii’s Tangible Media Group on the Tangible Co-Ideation project. The research examines embodied prompting interfaces for creative thinking with Large Language Models, extending his HCI work toward new forms of interaction with generative AI.
During his PhD, he also contributed to teaching in the Creative Technology major at the Institute for Future Technologies, mentored student projects and took part in several collaborative research projects.
His work received recognition at the MIT Hard Mode: Hardware AI Hackathon, where his IPheromone project won the Connect track.
A thesis supported by several research institutions
The thesis involved researchers and institutions from engineering, neuroscience and human-computer interaction. Olivier Lambercy and Frederic Bevilacqua served as reviewers, while Agnès Roby-Brami, Sabrina Panëels and Samuel Bottani joined the thesis jury.
With the PhD completed, Paul-Peter Arslan is pursuing research at the intersection of AI, embodied interaction, HCI, and neurorehabilitation, focusing on technologies that connect computational methods with human movement and interaction.
Learn more about ESILV’s research strategy
This post was last modified on 1 September 2026 4:56 pm