Fisher-HHT: A Feature Extraction Approach For Hand Gesture Recognition With a Leap Motion Controller
Nahla Majdoub Bhiri, Safa Ameur, Imen Jegham, Mohamed Ali Mahjoub, Anouar Ben Khalifa · 2022
In the last decades, Hand Gesture Recognition (HGR) has become one of the most prominent research topics, due to its wide range of applications in computer vision. Various approaches and techniques were suggested and evaluated to conduct significant results on HGR. However, even with the triumph reached on state-of-the-art methods, they have not considered the non-linearity and non-stationarity existing on time-series data including the Leap Motion Controller (LMC) raw data. In this paper, we propose a novel method for extracting and selecting pertinent and discriminant features using Hilbert Huang Transform (HHT) and fisher discriminant analysis. Therefore, the time-series signals of LMC are decomposed through an empirical mode decomposition. Then, the HHT is applied to generate the resultant Hilbert marginal spectrum. Next, the fisher discriminant analysis is proceeded for an efficient feature selection. Experimental results demonstrate the effectiveness of the proposed approach in term of hand gesture classification accuracy on the challenging LeapGestureDB dataset.