| Summary: | 博士 === 國立臺灣科技大學 === 資訊工程系 === 98 === Sign language usually requires large-scale movements to form a gesture, such as 3D maneuverings of two hands, palms and sometimes even arms, and thus more often than not presents varying degrees of difficulty among disabled people. In contrast, finger language is represented by small-scale hand gestures accessible by a mere change of the bending manner of a patient’s fingers. This thesis aims to develop an intelligent communication system to help the handicapped aphasiacs, who can only perform tiny finger movements, so that they can properly interact with the outside world. To develop such a communication system, we need to deal with both input hardware and recognition software.
For hardware, we have developed a new data glove which allows the light from light-emitting diodes (LED) to reach photo-detectors (PD) on line of sight. This new data glove possesses merits of better sensitivity, repeatability, and adjustability (or programmability).
For software, we have integrated fuzzy set theory and neural network to create two hybrid models for performing finger language recognition. The first model intended to support the basic communication requirement of seriously disabled patients by recognizing static finger gestures. The second model was used as a general communication system for handicapped aphasiacs to conduct continuous finger language input.
Our experiments show that the correctness of the geometric model of the data glove is verified by the experimental outcomes; thus, by adjusting the LED current, we can control the signal range of the data glove to adapt to a specific user. As for static finger language recognition, the user-dependent easily-adjusted static finger language recognition system has achieved 100% correctness for given specific users under unbiased field experiments. Finally, for continuous finger language recognition, our system can work well to perform basic communication tasks. That is, it can serve as a viable tool for natural and affordable communication for handicapped aphasiacs.
The contributions of the work can be summarized below. First, we have developed a programmable data glove aimed to assist the handicapped aphasiacs whose fingers can barely perform even very limited movements, such as bending. The data glove is sensitive, repeatable, and adjustable. In addition, it is mechanically simple, low-cost, and robust and thus can meet practical needs. Second, we have successfully developed a static finger language recognition tool for seriously disabled patients. The disabled can use the system to perform predefined finger gestures to express their intentions and resume their basic communication ability with other people. This is a great help for them to improve the quality of life. Finally, we have successfully developed an intelligent system for general communication of the handicapped aphasiacs. The system provides a suitable environment for them to perform continuous finger language input. Aside from working as a daily communication tool, it can support more advanced applications in the digital world. This can play a very significant role in further improving the life quality as well as the career development of the handicapped aphasiacs.
|
|---|
