Software-Defined Networking (SDN) has revolutionized network management through its centralized control architecture, but this centralization makes it particularly vulnerable to Distributed Denial of Service (DDoS) attacks, especially in IoT-enabled smart home environments. Existing detection approaches face challenges including dataset inconsistency, and time-efficient performance metrics. This paper proposes an Enhanced Ensemble Learning with Voting (EL-V) framework for efficient DDoS attack detection in SDN environments. The framework integrates optimized machine learning pipelines with ensemble techniques, emphasizing real-time applicability as well as dataset consistency. Using three distinct datasets (InSDN, UNSW-NB15_1, and CICIDS2017), our approach achieved remarkable performance: 94.84% accuracy and 94.99% precision on InSDN, 98.2% accuracy and 98.05% precision on UNSW-NB15_1, and 90.44% accuracy and 91.43% precision on CICIDS2017, with processing times under 0.34 s. The model consistently outperformed traditional machine learning approaches, K-Nearest Neighbors (KNN), Naive Bayes, Decision Tree and Logistic Regression, across all metrics, demonstrating its effectiveness for DDoS attack detection in SDN-IoT time-critical environments. These results suggest that EL-V offers a promising solution for enhancing network security in IoT-enabled smart homes while maintaining computational efficiency.