Facial emotion recognition (FER) has achieved significant advances in recent years, particularly for adult subjects. However, recognizing emotions in children remains challenging due to limited annotated data and notable differences in facial characteristics between adults and children. Models trained on adult datasets often underperform when applied to children, highlighting the need for effective domain adaptation. In this study, we propose a novel approach that combines transfer learning with multiple domain adaptation techniques to improve FER performance in children by leveraging adult facial emotion datasets. Specifically, we employ pre-trained models on adult data and strategically integrate several domain adaptation methods to address both data scarcity and domain discrepancy. Experimental results demonstrate that our integrated approach significantly enhances recognition accuracy in children. These findings suggest promising applications in child-centered domains such as education and healthcare, and contribute to advancing cross-domain FER techniques for more inclusive human-computer interaction.