Design and Implementation of Human Partner System with Sign Language Recognition and Visual Sensing Functions

Abstract

This dissertation constructs the HPS and introduces four sub-systems the hearing system speaking system sign language recognition system and vision system The hearing system utilizes the proposed voice detection and segmentation method to extract the meaning part of speech The speaking system allows the HPS to read the webpage read the new email and talk to the person In the visual sensing system besides the fundamental functions face detection and face recognition the partial face detection method also has been proposed and the CV value is introduced to appraise the degree of importance of the features Besides to allow the HPS to search for a target object and identify what this object is the DPCAMSHIFT algorithm and the anthropomorphic learning process have been proposed to achieve these goals For the hearing-impaired people the sign language recognition system for home-service-related TSL words is also implemented in this dissertation so that HPS can assist them in their daily lives The sign language recognition system utilizes HMM to classify the sequential data However there are two issues to be solved in HMM One is the number of states and the other is the determination of structure of HMM To determine the number of states in HMM this dissertation proposes the Entropy-Based K-means algorithm to plot the entropy diagram which provides a visualization method to judge the number of clusters for real datasets Finally this dissertation utilizes the ABC algorithm to establish the structure of HMM

Date of Award	2015 Jul 10
Original language	English
Supervisor	Tzuu-Hseng S. Li (Supervisor)