Weight-related issues have become a major global public health challenge, with obesity-related diseases imposing significant physiological, psychological, and economic burdens on individuals. Dietary behavior, as the primary pathway for calorie intake, plays a crucial role in weight management. Among these behaviors, eating speed significantly impacts energy intake and metabolic health—rapid eating delays satiety, leading to overconsumption, and is commonly observed among overweight populations. However, existing monitoring methods heavily rely on self-reporting or laboratory equipment, limiting real-time application in everyday life scenarios.This study proposes a real-time dietary behavior monitoring system utilizing audio signals from earphones. By collecting a dataset of chewing and swallowing sounds in both laboratory and real-world dining environments, the system detects chewing and swallowing events in natural settings, estimates eating speed, and triggers real-time auditory interventions when rapid eating is detected. The algorithm integrates audio energy features with deep neural networks, innovatively introducing a heuristic swallowing detection method based on chewing intervals, and develops a fully functional Android application to support real-time feedback. To enhance intervention effectiveness, four distinct styles of voice reminder schemes were designed and evaluated.