Indian Sign Language (ISL) interpretation is the major research work going on to aid Indian deaf anddumb people. Considering the limitation of glove/sensor based approach, vision based approach wasconsidered for ISL interpretation system. Among different human modalities, hand is the primarily usedmodality to any sign language interpretation system so, hand gesture was used for recognition of manualalphabets and numbers. ISL consists of manual alphabets, numbers as well as large set of vocabulary withgrammar. In this paper, methodology for recognition of static ISL manual alphabets, number and staticsymbols is given. ISL alphabet consists of single handed and two handed sign. Fourier descriptor as afeature extraction method was chosen due the property of invariant to rotation, scale and translation. Truepositive rate was achieved 94.15% using nearest neighbourhood classifier with Euclidean distance wheresample data were considered with different illumination changes, different skin color and varying distancefrom camera to signer position.
展开▼
