A Hybrid Method for Gesture Recognition

In this article, we propose a new hybrid approach with Fourier technique well known as Fourier descriptor and HU moments to detect hand gesture signatures for vision-based applications. Fourier descriptors are efficient specifications of image feature extraction. They can be used for shape characterization by preventing the scale, translation, and rotation invariance, as well as their independent from small shape deviations. To obtain good results for hand images, they converted into HSV color space. HSV unlike RGB color space has some features help users understand the nuances in the images; since, if one color be similar to another color, HSV can get better results. After that, Hand image will be extracted from HSV image by implementing the threshold operation. The Fourier Descriptors of extracted hand images are hybridized with HU moments and are compared with the reference dictionary to perform gesture recognition. Experiments are performed on the Bochum Gestures database. Finally the performance of the gesture recognition using Euclidean distance investigated. Results show that proposed hybrid method gives a high performance for hand gesture applications.

Hand gesture, HSV Color space, Hybrid Fourier Descriptor, Geometric Invariant Moments.