This paper presents a machine-learning-based method for detecting smart ring wearing actions using transient photoplethysmography (PPG) signals. During wear, the PPG sensor captures a distinct temporal pattern characterized by rapid amplitude fluctuations, irregular peaks, and a sharp transition to stable waveform. This dynamic signature is leveraged to distinguish genuine wear from accidental touches or partial contact with inanimate objects. To enhance power efficiency, a dual-sensor triggering mechanism—comprising a low-power accelerometer and an infrared (IR) proximity sensor—activates the PPG sensor only when wear likelihood is high. A dataset of 1,338 samples was collected from 104 human participants and 10 common inanimate objects (e.g., marker pen, lipstick tube, smart ring case) to represent realistic false-positive scenarios. Multiple models were evaluated, with Random Forest achieving the highest accuracy (94.1%) and Gradient Boosting performing comparably (93.4%), while LSTM showed lower performance (91.2%). The proposed approach improves wear-state reliability, reduces false detections, minimizes unnecessary sensor activation, and extends battery life, making it suitable for next-generation smart ring health monitoring systems