This paper proposes a new method that can recognize a sequence of hand motion expressing a sentence in sign language.
Recognition procedure is divided into two steps: separation of the sequence of hand motions into the sub-sequences each
of which expresses one word and combination of the words in order to construct a sentence having a meaning. In the first
step, sequences of hand motion images are segmented by testing the continuity of the hand motions and by the multiscale
image segmentation scheme. The trajectory of the hand motions are estimated by the affine transformation. Each
sign in the sentence is represented by the extended chereme analysis model and each chereme is represented by the status
vector for determining the transition in the HMM. In the second step, each sentence is also represented by the HMM. The
Viterbi algorithm and context-dependent HMM are used to find the best state sequence in the HMM. The proposed
algorithm has been tested with ten sequences of images, each of which expresses a sentence in Korean sign language.
The experimental results have shown that the proposed algorithm can separate the sentence level image sequence into the
word level sub-sequences with the success rate of 75% on average and recognize the sentence with the success rate of
This paper proposes a new fractal image encoder using a SOFM neural network based classifier and also an improved isometric transformation, to reduce the encoding time. Here the sizes of a domain block and range block are 8x8 pixels and 4x4 pixels, respectively. Block is classified into one of four patterns, based on the variation of intensities of the pixels in the block: flat where it is very low, middle where it is small, vertical/horizontal where there exists a vertical or horizontal edge, diagonal where there exists a diagonal edge. The SOFM neural network memorizes these patterns by competitive learning where the weights on the connections are determined by the Kohenen's learning rules. To reduce the searching time, the proposed algorithm searches domain blocks following a spiral trajectory starting from the block selected in the range and uses an improved isometric transformation which classifies the templates before comparison. The experimental results have shown that the proposed algorithm reduces the encoding speed by 50% on average while maintaining the same PSNR and bit rate, compared to the other's recent research results.
Since the shape of a 3D object moving in 3D space changes a lot in 2D image due to translation and rotation, it is very difficult to track the object using the SSD algorithm which finds the matching object in the input image using the template of the moving object. To solve the problem, this paper presents an enhanced SSD algorithm which updates the template based on an extended snake algorithm adaptive to the shape variation. The proposed snake algorithm uses the derivative of the area as the constraint energy to determine the boundary of an interested area considering the progressive variation of the shape. The performance of the proposed algorithm has been proved by the experiments where a mobile robot with one camera tracks a 3D target object translating and rotating arbitrarily in the 3D workspace.