A newly developed framework named CogniWave integrates contactless frequency-modulated continuous-wave millimeter-wave radar with an uncertainty-aware cognitive decision module to deliver privacy-preserving and safe intelligent automation in everyday environments. Detailed by researchers in the Journal on Advances in Signal Processing, the system pairs contactless radio-frequency sensing with cognitive artificial intelligence to overcome persistent operational challenges in human-centered automated monitoring.
Reliable human-centered automation has historically proven difficult to deploy across routine domestic and workplace settings. Conventional camera-based sensing systems continually raise serious privacy and bystander-consent concerns, whereas wearable-based monitoring relies heavily on long-term user compliance to maintain effectiveness. By relying entirely on contactless millimeter-wave radar hardware, CogniWave avoids capturing visual camera streams while bypassing the compliance burdens of body-worn devices.
The proposed architecture is organized into three distinct tiers designed to transform raw sensor measurements into controlled responses. At the initial level, a Perception Layer converts raw radar in-phase and quadrature (IQ) signals into compact, machine-consumable feature representations. These extracted parameters specifically include a micro-Doppler motion embedding, three-dimensional trajectory kinematics, a respiration variability index (RVI), and a dedicated radar reliability score.
At the intermediate stage, the Cognitive Interpretation Layer processes these compact feature sets using a specialized network structure termed CogiNet. This layer maps the incoming kinematic and physiological representations into categorical activity probabilities, structured human-state tokens, and an explicit uncertainty signal. By quantifying operational uncertainty directly from radar-derived features, the network supplies crucial probabilistic context for downstream automated actions.
The third tier consists of a Human-Centered Automation Layer that enforces an explicit act, abstain, or clarify policy to govern system interventions. This layer prioritizes user safety and autonomy by integrating user control features, including manual override and opt-out options, alongside systematic transparency logging for system actions.
To evaluate everyday interactions, the system focuses on an eight-class activity set: empty, presence-still, sitting, standing, walking, sit-to-stand, fall, and inactivity anomaly. According to the research team, these eight target categories were explicitly selected to achieve a practical balance between routine daily automation triggers and safety-critical monitoring events.
The experimental validation evaluated system robustness across both subject-disjoint and environment-disjoint data splits. The complete experimental dataset encompassed 4,000 trials collected from 50 participants, representing 328 minutes of millimeter-wave radar recordings. From these gathered sessions, the investigators processed 37,128 segmented analytical windows to systematically benchmark performance across unfamiliar individuals and varied physical environments.
Across these experimental tests, CogniWave demonstrated robust performance, achieving 94.1 percent overall accuracy alongside a 0.93 macro-F1 score. For safety-critical situations, the framework maintained a high event recall of 0.98 for detected falls, while restricting the rate of false triggers to 0.6 false alarms per hour.
The framework further incorporates reliability-aware gating, which actively reduces false triggers during degraded sensing conditions and sustains low selective risk at practical coverage levels. Finally, the authors evaluated real-time feasibility on embedded hardware platforms, achieving sub-second end-to-end latency per window and confirming that the radar-based framework can support online deployment in everyday operational settings.
Subject of Research: Contactless millimeter-wave radar and cognitive AI for human activity recognition
Article Title: CogniWave: integrating cognitive AI and mmWave radar for human-centered intelligent automation
Article References: Elinjulliparambil, S. H., Dalal, P., Ali, T., Gumber, S., Imran, N., & Choi, G. S. (2026). CogniWave: integrating cognitive AI and mmWave radar for human-centered intelligent automation. Journal on Advances in Signal Processing. https://doi.org/10.1186/s13634-026-01380-6
Image Credits: AI Generated
DOI: 10.1186/s13634-026-01380-6
Keywords: millimeter-wave radar, cognitive artificial intelligence, human-centered automation, FMCW radar, activity recognition, fall detection, CogiNet, micro-Doppler motion embedding, 3D trajectory kinematics, respiration variability index, uncertainty awareness, contactless sensing, inactivity anomaly, reliability-aware gating, embedded hardware, selective risk, signal processing, privacy-preserving sensing
Cite Scienmag News
Blake Davidson. (November 16, 2026). Cognitive Radar Framework Accurately Detects Human Activity and Fall Events. Scienmag. https://scienmag.com/cognitive-radar-framework-accurately-detects-human-activity-and-fall-events/
Blake Davidson. "Cognitive Radar Framework Accurately Detects Human Activity and Fall Events." Scienmag, 16 November 2026, https://scienmag.com/cognitive-radar-framework-accurately-detects-human-activity-and-fall-events/. Accessed 17 November 2026.
Blake Davidson. "Cognitive Radar Framework Accurately Detects Human Activity and Fall Events." Scienmag. November 16, 2026. https://scienmag.com/cognitive-radar-framework-accurately-detects-human-activity-and-fall-events/








