Assessment of Swallowing Sounds' Stages with Hidden Markov Model

Swallowing sounds analysis has received much attention recently due to its diagnostic potentials. The process of swallowing is one of the most complicated mechanisms of the body, in which a few different events are occurred in a few milliseconds. The timing and coordination of these events within a breath cycle is crucial for a safe swallow to occur. In order to improve our ability to treat swallowing disorder and to create non-invasive acoustical diagnostic methods, we must develop a detailed understanding of how the swallowing sounds are produced and what the contribution of each swallowing event is to the sounds that are heard and recorded as swallowing sounds. This paper presents a Hidden Markov Model (HMM) of the swallowing sounds, in which it is shown for the first time that the swallowing sounds have three stages. The number of stages was found to be very consistent between the subjects and the bolus textures implying the robustness and fitness of the model.