In-ear acoustic monitoring for swallow detection

Sialorrhea, or involuntary drooling, affects up to 25% of patients with neurodegenerative diseases, causing significant discomfort. Current treatments are mostly invasive. Non-invasive treatment may be possible by providing patients with reminders to swallow based on their swallowing frequency. A promising approach involves detecting swallowing using in-ear devices. To enable such treatment, we are collecting a database of 13 non-verbal human-produced events such as blinking, coughing, grinding teeth, etc. These events are captured by various sensors inside the ear, including microphones, air pressure microphones, PPG sensors, and IMUs, in different acoustic environments, both quiet and noisy. Additionally, we use ultrasound videos of the tongue as a ground truth for spontaneous swallowing. We will use this database to train machine learning algorithms for detecting and classifying swallowing and other non-verbal events. This work will enable the development of health monitoring in-ear devices for vulnerable individuals.