Real-time, continuous, and non-invasive blood pressure (BP) monitoring is of great importance for the early detection and long-term management of hypertension. With the rapid development of wearable technologies, signal acquisition methods based on photoplethysmography (PPG) and micro-cuff oscillometry have gradually emerged as engineeringoriented alternatives to conventional cuff-based BP measurement. In recent years, artificial intelligence-based BP estimation algorithms have proliferated, whose performance depends on factors such as signal quality, feature construction, and model architecture. This paper provides a systematic overview of the key technical pipeline for wearable BP estimation, including signal acquisition methods, signal feature extraction, model development, and performance evaluation criteria. It focuses on a comparative analysis of pulse transit time (PTT)-based methods, PPG feature-machine learning methods, PPG signal-deep learning methods, and micro-cuff oscillometry, and further summarizes recent algorithmic advances based on emerging wearable BP measurement technologies, including ultrasound, bioimpedance, and tonometry. Furthermore, this paper summarizes the major challenges in the field, including limited model generalizability, noise and motion artifacts, weak adaptability to individual variability, dependence on initial calibration, and concerns regarding data privacy and ethics. Potential future directions are also discussed, such as multi-feature fusion with mechanism-constrained modeling, few-shot personalized learning, robust temporal modeling for long-term continuous monitoring, unified multimodal modeling with system co-optimization, and standardized evaluation with clinical adaptability. This review aims to provide technical insights and development recommendations for the measurement technology research, algorithm design, and engineering application of wearable BP monitoring systems.