Traditional human computer interaction methods often rely on physical devices such as mice or touchpads, which limit accessibility and naturalness in immersive or contactless environments. This paper proposes AeroCursor, a novel adaptive air-gesture–based cursor control system that integrates spatiotemporal neural mapping with depth-aware sensor fusion to achieve intuitive and precise on-screen navigation. Unlike existing systems that depend solely on 2D image processing or static gesture recognition, AeroCursor continuously learns user specific motion dynamics by capturing temporal gesture evolution in 3D space through a fusion of vision, ultrasonic, and inertial data streams. The system further incorporates a context-aware intent prediction model that forecasts probable cursor destinations using gaze and movement priors, significantly reducing pointer drift and latency. Experimental results demonstrate over 92% tracking accuracy and 30% faster target acquisition compared to conventional gesture-based systems. The proposed architecture opens new possibilities for touchless computing, AR/VR interfaces, and assistive technologies, offering a more natural, adaptive, and intelligent human–machine interaction paradigm.