Nocturnal enuresis is a common issue among toddlers, young children, and older adults, often caused by the inability to wake up before bladder overflow. While children typically develop waking habits over time, bedridden older individuals depend on caregivers, increasing the risk of delayed intervention. Existing enuresis alarms struggle with false positives, as they cannot reliably distinguish between sweating and urination. Early resistive-based alarms suffered from current flow through electrolytes, while modern capacitive-sensing alarms, using two-electrode designs, still fail to differentiate moisture sources accurately. This research focuses on developing an advanced embedded system for enuresis alarms, utilizing a microcontroller and enhanced capacitive sensing, combined with algorithms trained using machine learning techniques. By analyzing the rate of moisture spread, intensity, and affected area, the system effectively distinguishes urination from sweating, improving detection accuracy. Experimental results indicate a significant reduction in false alarms. Future work will explore dynamic calibration, improved response time, and additional performance enhancements.