Tiny Machine Learning (TinyML)–an ultra-lightweight machine learning technology deployed on edge devices is becoming a key solution for innovative wearable applications such as health monitoring and gesture recognition. This paper provides a systematic overview and in-depth quantitative analysis of research works applying TinyML in wearable application development from 2020 to April 2025. Using the PRISMA standard review method combined with bibliometric analysis, the study selected 59 representative works from 954 documents collected from Web of Science, Scopus, IEEE Xplore, ACM, MDPI, and arXiv. The analysis identifies three principal development axes (model, hardware, implementation), classifies 10 common challenge groups, and maps them to potential research directions, including federated learning, lightweight feature extraction, and multi-objective simultaneous optimization. Additionally, it examines publication growth trends, concept evolution, and prominent keyword phrases. The findings offer a comprehensive academic perspective and a strategic roadmap for researchers developing TinyML applications on wearable devices.