Gesture Recognition using Dynamic Movement Primitives

Please enable javascript in your browser.

Asemanrafat, Amirreza | 2023

66 Viewed

Type of Document: M.Sc. Thesis
Language: Farsi
Document No: 55979 (08)
University: Sharif University of Technology
Department: Mechanical Engineering
Advisor(s): Taheri, Alireza; Meghdari, Ali
Abstract:
In this thesis, we introduced a new augmentation method that takes into account the inherent properties of trajectory data and regenerates valid trajectories while preserving all the distinctive features of the main path. Our method uses Dynamic movement primitives (DMP) formulation, which is widely used in path generation in robotics, to manipulate the data in a kinematically accurate way. We implemented the presented method on our Iranian sign language data set by augmenting each group in our data set with a proper form of our DMP data augmentation method. After training our augmented data set with two deep classification models, We achieved 82.95 percent maximum and 77.61 percent mean accuracy in one-shot learning, and 86.2 percent maximum and 81.86 percent mean accuracy with two data from each class. On its own and combined with other methods, our method can improve the classification and prediction results in various types of trajectory data sets
Keywords:
Dynamic Movement Primitives ; Data Augmentation ; Sign Language Recognition ; One-Shot Learning ; Skeleton Action Recognition

No TOC