Research
Ethan AI is built at the intersection of behavioral understanding, computer vision, and assistive systems. We track the research that informs our work, and we collaborate with practitioners to turn it into deployable, real-world monitoring—inside therapy rooms and across fixed CCTV setups.
CATEGORIES
Wandering & Elopement Distress & Escalation Aggression & Peer Safety Drooling & Medical Risk Self-Harm & High Risk ADLs & Dressing Therapy / Classroom Cues Outside-Camera Safety CollaborationsWandering & Elopement
Research that supports detection of boundary-crossing, exit intent, unsupervised movement, and “missing child” risk patterns using fixed cameras.
Paper Title (Replace)
1–2 lines: what the paper does, and why it matters for wandering / elopement detection.
ReadPaper Title (Replace)
Short note: method + outcome relevant to elopement / boundary alerts.
ReadDistress & Escalation
Work on agitation, dysregulation, repetitive motion, posture changes, pacing, and early signs of escalation—signals caregivers often miss under load.
Paper Title (Replace)
1–2 lines: what the approach detects (movement/posture/audio proxies) and how it relates to escalation.
ReadAggression & Peer Safety
Research around detecting contact events (pushing, pinching, hair pulling), unsafe proximity, and conflict onset in group environments.
Paper Title (Replace)
Short note: how it models interactions (two-person tracking, pose, contact cues) and reduces caregiver load.
ReadDrooling & Medical Risk
Work relevant to face/mouth region cues, salivation proxies, and short-sequence detection where labeled data is typically scarce.
Paper Title (Replace)
Short note: technique + dataset style; relevant to drooling and adjacent care cues.
ReadSelf-Harm & High Risk
Research on high-risk patterns: head banging, biting, hitting self, choking-risk precursors, and other safety-critical cues.
Paper Title (Replace)
Short note: what it detects, false positive control, and why it’s safety-critical.
ReadADLs & Dressing
ADL-oriented research (dressing compliance, hygiene cues, task progress). This category is often constrained by privacy and consent-heavy data pipelines.
Paper Title (Replace)
Short note: task understanding / activity recognition approach applicable to dressing-related cues.
ReadTherapy / Classroom Cues
Research that informs what fixed cameras can capture during therapy: engagement cues, withdrawal, posture, pacing, and safety-relevant events.
Paper Title (Replace)
Short note: relevance to classroom/therapy monitoring and reporting.
ReadOutside-Camera Safety
Relevant work for perimeter cameras: exit events, road-side risk, unsafe wandering paths, and unattended child detection.
Paper Title (Replace)
Short note: perimeter safety, tracking robustness, and alert thresholds for outside cameras.
ReadCollaborations
Our research-to-reality path depends on collaboration: therapists, special schools, and partners who help us validate systems ethically in the real world. We also track and support research aligned with our mission, while building deployable behavior-recognition systems on top of fixed cameras.
• “Our work is informed by peer-reviewed research and validated through real-world deployments.”
• “We collaborate with practitioners to translate research into deployable behavior recognition.”
• “We support and follow research aligned with our mission, and we build systems that implement these ideas responsibly.”
Tip: For each paper, add a one-line “Why it matters” in caregiver language (not research language).