The Accuracy of the k-Nearest Neighbors and k-Means Algorithms in Gesture Identification

Guzsvinecz, Tibor and Szűcs, Judit and Szücs, Veronika and Demeter, Róbert and Katona, József and Kővári, Attila (2024) The Accuracy of the k-Nearest Neighbors and k-Means Algorithms in Gesture Identification. INFOCOMMUNICATIONS JOURNAL : A PUBLICATION OF THE SCIENTIFIC ASSOCIATION FOR INFOCOMMUNICATIONS (HTE), 16 (SI). pp. 30-36. ISSN 2061-2079

Preview

Text InfocomJournal_2024_SpecISS.pdf - Published Version Download (1MB) | Preview

Abstract

In today’s digital era, human-computer interac tion interfaces evolve and increase together with the needs of the users. However, the existing technologies have their limita tions, which can hinder the efficiency of modern input devices like the Kinect sensor or other similar sensors. In this paper we improved our previous algorithm by extending it with two algo rithms that aim to help telerehabilitation for individuals with movement disabilities. These two algorithms are based on the k Nearest Neighbors, the k-Means algorithms. The algorithms are designed to accommodate the needs of the patients by adapting to their gestures based on their previous three. Using these ges tures, the algorithms create multiple gesture acceptance domains around each coordinate of the gesture. Consequently, they decide whether the next user-input gesture can be considered the same movement. The accuracy of these algorithms was evaluated in three acceptance domains by comparing gesture descriptors with either the Euclidean or the Manhattan distance calculation meth ods. The results show that k-Nearest Neighbors algorithm yields better results in larger acceptance domains, while the k-Means algorithm can provide a better gesture acceptance rate in the smaller ones. The results show that both algorithms can be used in the telerehabilitation process, although the k-Means algorithm is more accurate than the k-Nearest Neighbors algorithm.

Actions (login required)

Edit Item