Edge Intelligence Systems: Optimizing Latency, Accuracy, and Resource Efficiency

Abstract

Edge intelligence systems have emerged as an effective solution for improving computational efficiency and reducing processing overhead in smart healthcare Internet of Things (IoT) environments by enabling localized data analysis at edge devices. This research proposes a Sterna Migration-Optimized Lightweight Convolutional Neural Network (SM-L-CNN) method to enhance edge intelligence performance in smart healthcare IoT systems. Healthcare data are collected from wearable sensors, patient monitoring devices, and publicly available medical IoT datasets containing physiological signals and clinical parameters. The collected data undergo preprocessing using min–max normalization to standardize feature values and ensure uniform data representation across heterogeneous sensor inputs. Feature extraction is performed using Discrete Wavelet Transform (DWT) to decompose physiological time-series signals into multi-scale frequency components and capture essential health-related patterns. The extracted features are processed using a lightweight convolutional neural network deployed at the edge layer, while the Sterna Migration (SM) optimization algorithm is employed to optimize network parameters, filter weights, and resource allocation to reduce computational overhead and latency. The proposed SM-L-CNN framework improves classification accuracy, minimizes processing delay, and enhances energy-efficient resource utilization under constrained edge environments. Experimental results demonstrate superior performance in latency reduction, prediction accuracy of 0.953, precision of 0.925, recall of 0.958, and F1-score of 0.914. Computational efficiency compared to conventional edge intelligence models, providing a scalable and intelligent solution for smart healthcare monitoring and edge-based medical IoT applications.