MINDS CDT doctoral researcher | University of Southampton
Teaching AI to understand sign language through video and radar
Keniel Peart is developing AI methods that interpret sign-language movement from video and radar, using 3D models of the human body and hands. His research could support more natural assistive communication tools and sensing that does not rely solely on cameras.
Profiles: LinkedIn | GitHub | Google Scholar
Research at a glance
| Research focus | Sign-language recognition and human-motion understanding |
| MINDS theme | Embedded Artificial Intelligence |
| Sensing | Video and radar |
| Technologies | Machine learning, computer vision, 3D human modelling, radar simulation |
| Potential applications | Assistive communication tools, privacy-aware sensing, human-computer interaction |
| Current status | Doctoral researcher |
| Programme | MINDS CDT, 2022 |
| Expected graduation | 2027 |
Understanding sign language through movement
Sign languages are expressed through movement. Meaning is carried by the hands, fingers and body working together, and a small change in a hand shape or movement can change what is being said.
AI systems that attempt to recognise or translate signing can find these fine movements difficult to capture. Methods that perform well in controlled research settings may be less accurate or less realistic in everyday environments, and may not generalise well to new people and places.
Most current approaches also depend on cameras. Cameras are not always suitable: lighting and visibility vary, and some settings call for sensing that reduces reliance on visual imagery for privacy reasons.
Keniel is developing methods that recover detailed 3D body and hand movement from video over time. Alongside this, his work simulates how sign-language movements would appear to radar, a sensing technology that detects motion using radio waves. Together, these strands support research into systems that can combine or compare information from different kinds of sensor.
Why the research matters
Supporting more natural assistive communication
More accurate understanding of hand and body movement could contribute to assistive communication tools that feel more natural for sign-language users. Keniel’s work focuses on developing and testing the methods that future tools could build on.
Sensing that does not rely only on cameras
Radar senses movement without recording a conventional picture of a person. Radar-based approaches could support sensing in places where cameras are poorly suited, or where reducing reliance on visual imagery may offer privacy advantages.
Wider relevance to human-computer interaction
Recognising detailed hand and body movement is also relevant to human-computer interaction more broadly, where people may interact with technology through gesture.
Research approach
Keniel’s research combines machine learning with computer vision, the field of AI that interprets images and video. From video, his methods build a 3D model of a person’s body and hands, estimating the position of joints and fingers (known as body and hand pose estimation), and track how that model moves over time.
He also uses radar simulation: taking recovered 3D movement and modelling the signal a radar sensor would receive if it observed the same motion. This offers a controlled way to study how radar responds to signing movements.
Video, radar and multimodal sensing
Video and radar capture movement in very different ways. Video records what a scene looks like, while radar measures how radio waves reflect from moving objects, so one is not a direct substitute for the other. Multimodal sensing means combining or learning from more than one kind of signal, so that the strengths of one could help offset the limitations of another. Keniel’s research investigates how these different signals can be compared and brought together to understand the same movement.
Research keywords: Sign-language recognition, sign-language translation, computer vision, 3D human modelling, body and hand pose estimation, radar simulation, radar sensing, multimodal sensing, machine learning
Research and outputs
Radar research for brain imaging
Keniel has also researched radar for a different application: non-invasive brain imaging.
Featured output: Advancements in non-invasive neuroimaging: exploring the potential of radar technology for brain imaging and tumour detection. Keniel Romario Peart, Shelly Vishwakarma and Indu Bodala. Presented at the 46th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC), Orlando, July 2024.
View the EMBC 2024 paper record in ePrints Soton
Research profiles
Collaboration and public engagement
DJ Air: mixing music with hand gestures
Keniel is one of the MINDS postgraduate researchers behind DJ Air, which lets people mix music and control lighting with hand gestures picked up by an ordinary laptop camera. He built it with fellow MINDS researchers Rudra Mutalik and Epifanios Baikas and SustAI CDT postgraduate researcher Theo Parpounas. DJ Air began as the winning idea at the hackathon during the 2025 MINDS and SustAI Innovation Camp, and went on to win Best Demonstration at the ECS Student Showcase.
Read how the gesture recognition behind DJ Air works
DJ Air was also demonstrated to the public at the British Science Festival 2026, where Keniel attended to see visitors try it. Read about MINDS researchers at the British Science Festival.
Looking ahead
After his PhD, Keniel hopes to work in industry on computer vision and robotics.
Connect with Keniel
Interested in sign-language technology, computer vision or radar sensing? Connect with Keniel through his LinkedIn profile, or contact the MINDS team about a wider research or collaboration enquiry.
To explore his work further, see Keniel’s code on GitHub and his publications on Google Scholar.