A method can use dual-axis accelerometry signals obtained during a swallow toclassify the swallow as a normal swallowor as an impaired swallow (e.g., an aspiration- penetration). The method caninclude representing the dual-axis accelerometry signalsas meta-features, comparing the salient time and frequency meta-features,identified by regularized binomial logistic regression withelastic net penalty performed on the time and frequency meta-features in aknown training data set, with a preset linear discriminantclassifier constructed based on the salient meta-features, and classifying theswallow as a normal swallow or a possibly impairedswallow, based on the comparing. Preferably a processing module operativelyconnected to the sensor performs the processing of thedual- axis accelerometry signals and also automatically classifies theswallow.
