The isolation of the deaf community from mainstream society is in part due to the lack of knowledge most hearing people have of sign language. To most, there seems to be little need to learn a language that is spoken by such a small minority unless perhaps a relative is unable to hear. Even with a desire to learn, the task may seem insurmountable due to the unique formational and grammatical rules of the language.
This linguistic rift has led to the call for an automatic translation system with the ability to take voice or written text as input and produce a comprehensive sequence of signed gestures through computing.
This thesis focused on the development of the foundation of a system that would receive English language input and generate a sequence of related signed gestures each synthesized from their basic kinematic parameters. A technique of sign specification for a computer-based translation system was developed through the use of Python objects and functions. Sign definitions, written as Python algorithms, were used to drive the simulation engine of a human-modeling software known as Jack. This research suggests that 3-dimensional computer graphics can be utilized in the production of sign representations that are intelligible and natural in appearance.
|