Accelerometer signal encoding using deep learning and spectrograms for swallowing classification
IEEE EMBS (In submission), 2024
Yuewen Luo; Ayman Anwar; Farnaz Khodami; James L. Coyle; and Ervin Sejdic
This work introduces spectrogram-based CNN and RNN encoders that compress high-dimensional HRCA signals into compact representations while preserving critical swallowing features, achieving state-of-the-art accuracy for dysphagia classification and supporting real-time bedside assessment.




