Integrated sensing and communication (ISAC) lets a base station detect human motion with the same signals it uses for data, but every sensing transmission takes resources from users. Sensing rarely while nothing moves saves resources, yet a brief hand gesture by an otherwise still person can then start and end between two transmissions and go unnoticed. We propose an evidence-driven schedule, based on data-efficient cumulative-sum (CuSum) change detection, that pauses sensing while measurements indicate no motion and senses continuously once they suggest motion. It is adapted to orthogonal frequency-division multiplexing (OFDM) sensing, where requests can be blocked by data traffic and measurements arrive late, and evaluated with an articulated human model of limb motion. In simulations without wall reflections, with the same monitoring budget and below one false alarm per minute, it detects 1.5--4.2 times as many limb movements within 200~ms as periodic sensing. A tuned two-rate trigger performs similarly, and the gain over periodic sensing persists, without retuning or cost matching, in a simulated room with wall reflections. Under a small budget, measurements should decide when to sense.

翻译：暂无翻译