No abstract

Several challenges like changes in brightness, dynamic background, occlusion and inconsistency of camera position make the recognition of hand gestures difficult in any vision-based method. Diversity in finger shape, size, distribution and motion dynamics is also a big constraint. This leads to the motivation in developing a dense Scale Invariant Feature Transform (SIFT) flow based architecture for recognizing dynamic hand gestures. Initially, a combination of three frames differencing and skin filtering technique is used for hand detection to reduce the computational complexity followed by a SIFT flow technique to extract the features from the detected hand region. SIFT flow vectors obtained from every pixel can lead to overfilling, data redundancy and dimension disaster. A dual layer belief propagation algorithm is utilized to optimize the feature vectors to resolve the dimensionality problem. Support Vector Machine (SVM) and K-Nearest Neighbors (KNN) classifiers are used to evaluate the performance of the developed framework. Experiments were conducted on hand gesture database for HCI, Sebastien Marcel Dynamic Hand Posture Database and RWTH German finger spelling database. The simulation results demonstrate that the developed architecture has excellent performance on the uneven background and varying camera position and it is robust against image noise. A comparative analysis with the state of the art methods illustrates the effectiveness of the architecture.

In this work a framework based on histogram of orientation of optical flow (HOOF) and local binary pattern from three orthogonal planes (LBP_TOP) is proposed for recognizing dynamic hand gestures. HOOF algorithm extracts local shape and dynamic motion information of gestures from image sequences and local descriptor LBP is extended to three orthogonal planes to create an efficient motion descriptor. These features are invariant to scale, translation, illumination and direction of motion. The performance of the new framework is tested in two different ways. The first one is by fusing the global and local features as one descriptor and the other is using features separately to train the multi class support vector machine. Performance analysis shows that the proposed approach produces better results for recognizing dynamic hand gestures when compared with state of the art methods

No abstract

No abstract