An E-health System Recognizing Vegetable Images Using Extreme Learning Machine

Zhenyu Wu, Yu Zhang, Yanqin Mao, Joel J. P. C. Rodrigues

Abstract: Smart devices are increasingly important in daily life as they can provide information and capture usage behavior. This paper proposes a machine learning-based system to assist in human health management on smart devices. The system architecture includes a data layer, function layer, and application layer with the goal of helping individuals identify healthful vegetables best suited to their dietary needs. The proposed system utilizes the Extreme Learning Machine (ELM) algorithm to accurately recognize vegetable images. Compared to deep learning techniques, ELM has more efficient training and inference processes, making it better suited for smart device applications. The experiment with the collected vegetable image dataset found that the relu activation function and Gaussian distribution weight initialization method yielded optimal performance for the proposed system. Additionally, ELM outperformed deep learning techniques with small amounts of data. A case study was implemented on the Android platform to demonstrate the feasibility of the proposed system.