To tackle the dual challenges of limited spectral resources and energy constraints in Internet of Things (IoT) networks, this paper proposes an innovative downlink non-orthogonal multiple access (NOMA)-enabled cognitive radio (CR) system integrated with simultaneous wireless information and power transfer (SWIPT). By jointly optimizing NOMA and SWIPT protocols, the proposed framework significantly improves spectral utilization while efficiently fulfilling the energy harvesting (EH) requirements of secondary users (SUs). Unlike prior works, we concentrate on the integrated optimization of spectral assignment, power allocation, and power splitting (PS) ratios under practical EH constraints, which poses a highly non-convex and coupled problem. To tackle this, we propose an iterative optimization framework: first, a low-complexity greedy method determines suboptimal sub-channel allocation; then, primal-dual alternating updates are employed to optimize power allocation and PS ratios. Numerical simulations reveal that our scheme substantially outperforms conventional orthogonal multiple access (OMA) and benchmark NOMA methods in terms of achievable data rate, demonstrating its superior capability to harmonize spectral efficiency with energy sustainability.