Person Identification from Contextual Motion

Researchers have introduced a novel interactive system that identifies individuals by analyzing their unique motion styles, using a sequential cue-and-response process inspired by human cognition. The method, detailed in a paper submitted to arXiv on 11 June 2026, frames person identification as a probabilistic message exchange between the subject and the system [1]. At each step, the system presents a visual stimulus and records the subject's motion response. The cue is chosen to maximize the mutual information between the expected response and the subject's identity, and the response updates the probability distribution over possible identities. The session ends once a sufficient classification confidence level is reached [1]. The authors state this is the first time person identification has been addressed in such an interactive setting [1]. The model is grounded in the Human Information Processing (HIP) paradigm, which treats the subject's behavior as a generative process [1]. This approach departs from traditional computer vision tasks such as activity recognition, object detection, and video tracking, which typically extract information from static or pre-recorded image sequences [4]. By making the identification session adaptive, the system can select the most discriminating cues for each individual, rather than relying on a fixed set of motions. To validate the approach, the researchers tested the model on five publicly available datasets and a newly collected dataset comprising 4,476 recordings of 22 test subjects responding to 15 cues [1]. The paper reports high recognition rates across these benchmarks, though specific numeric performance metrics were not detailed in the abstract [1]. The interactive nature of the protocol means the number of cues presented can vary per subject, terminating only when the system's confidence threshold is met. The work sits at the intersection of computer vision and spatial contextual awareness, where a system synthesizes information about a person's location, activity, and proximity to make decisions [5]. While motion graphic design focuses on the aesthetic combination of design and animation for media [3], this research treats motion as a biometric signature. The arXiv submission is hosted under the Computer Vision and Pattern Recognition category and is supported by the arXivLabs framework, which enables community collaborators to develop and share new features on the platform [1].