IEEE Access (Jan 2024)
Vibration Signal Based Abnormal Gait Detection and Recognition
Abstract
Experts have been researching different types of gait since the 19-century. The way people walk can give a myriad of clues as to the health of a person. Abnormal gait detection might help protect senior people from injury and reveal underlying health problem. In aging societies, the application of recognition of abnormal gait based on vibration signal is very useful, especially for those people who live by them own. Unlike other methods in the related research requiring image acquisition equipment and wearable device to identify relevant feature information, not even mention many are intrusive or too complicated for users. The proposed systematic prototype firstly uses foot vibration signals as the source for abnormal gait and fall detection. This paper investigates algorithmic aspects; the particular algorithm’s framework involves gathering data from several artificial sensors. An altered version of the Dynamic Time Warping (DTW) algorithm computes the anomaly index after splitting the active portion into active elements and denoising the active elements. Next, the K-Nearest Neighbor (KNN) algorithm separates the anomaly indices into distinct groups and generates the projected values representing the user’s gait. Ultimately, the predicted values are processed by the Hidden Markov Model (HMM), which then determines the user’s gait. In the meantime, if an abnormality of gait arises for various experimental environments, subjects, and shoe types, etc., its corresponding index and the value representing the user’s gait will also change in comparison to his or her normal gait, which will remain unaffected by the environment and the shoe type of the subject. As a result, the experiments in this paper are flexible. In the experiments, various sensor placements and subjects can also, to the greatest extent possible, reflect the algorithm’s adaptability to these changes.
Keywords
