This paper explores the intersection of Tiny Machine Learning (TinyML) and airwriting technique, focusing on their potential to empower individuals with visual and hearing impairments in Vietnam. TinyML offers efficient, real-time machine learning on low-power devices, making it ideal for applications in accessibility and assistive technology. While prior works often transform 3D motion data into image-based representations for character detection—resulting in additional processing time and resource consumption—this research proposes a direct feature-based classification method.