Human Activity Recognition (HAR) utilizing wearable sensors is widely applied in healthcare, industry, and daily life. This study explores the application of Transformer-based models, specifically the Vision Transformer, for HAR tasks, offering an alternative approach to conventional time series analysis and earlier deep learning methods. In this paper, we propose a model named ViT-HAR and evaluate its performance on the benchmark HARTH dataset. Experimental results demonstrate that ViT-HAR achieves competitive recognition performance across several metrics, providing a robust solution for HAR. This work highlights the potential of Transformer-based models for facilitating non-intrusive, real-time activity recognition across diverse domains.