Snoring is defined as sounds made by vibrations in the soft palate and their adjacent tissues during sleep. Heavy snoring can result in sleep-related upper airway narrowing, which leads to respiratory flow limitation and increased respiratory effort. If untreated, heavy snoring may be complicated by excessive daytime sleepiness. Hence, snoring has received a great deal of clinical attention in recent years. We identify the snoring sources and predict the snoring noise levels for a 3D human head model. Our human head model includes the upper part of head, neck, the soft palate, hard palate, tongue, nasal cavity and the surrounding walls of the pharynx. The snoring mechanism is investigated by applying the concept of structural intensity to a 3D finite element model of a human head. Results demonstrated that the vibrations of tissues are mainly in the areas of soft palate and tongue and nasal areas under fluid flow loading. For predicting snoring noise level, a 3D boundary element cavity model of upper airway in snoring is generated. The snoring noise level is predicted for the airflow loading estimated from the simulation of the fluid-structure interactions, and its range agrees with published measurements. These models may be further developed to study the various snoring mechanisms for different groups of patients. In future work, other details such as extrathoracic airways and fluid-structure interactions will be considered.