Sign Language Glove Prepared By Jaffer Kapasi Chris Lee Michael Mah Dave Yeung
Overview Based on the American Sign Language Translates hand positions into letters displayed on a screen Allows finger-spelling of words
Purpose Breakdown communication barriers between speech-impeded people and the general public Allow deaf/dumb people to communicate without a translator
Function Training Phase Training Program trains the SLG Translation Phase Translates hand positions into letters displayed on an LCD screen
SLG Design Diagram
FPGA Board Nios Development Kit FPGA APEX EP20KE200EFC484-2X
Glove Sports glove 6 flex sensors (5 finger, 1 wrist) Variable resistance sensors measures the amount of flex 2 contact sensors Determines contact between thumb/index and index/middle fingers
Analog To Digital Converter National Semiconductor ADC0809 Converts the variable analog voltage from the glove to digital values used by the Training Unit and Translation Unit
RAM Controller LPM RAM Stores the ‘base’ voltage values for each flex and contact sensor for each letter
LCD Lumex LCD LCM-S01602DSR/F-Y Displays the output of the SLG program
Training Unit Allows the SLG to be customized for individuals Initializes the ‘base’ voltage values in memory
Training Unit Algorithm Display on LCD “LETTER?” User signed “LETTER ” ? Move on to next letter Start All letters complet e? yes no Get flex sensor values and contact sensor values Store values to memory Done
Translation Unit Translates the hand position into a letter in the alphabet Compares the ‘base’ (in memory) and ‘measured’ SLG voltage values to obtain the closest matching letter
Translation Unit Algorithm Obtain ‘measured’ and ‘base’ values Contacts compare ? Move on to next letter Find the letter with the smallest ‘absolute difference’ Output letter to LCD Start Sum the absolute differences between the ‘measured’ and ‘base’ values for each flex sensors 1 to 6 If ‘measured’ and ‘base’ flex sensor value greater than 0.7 Volts then remove letter from consideration All letters complet e? yes no Remove letter
Achievements ~90% overall accuracy
Issues Translation accuracy very dependent on user Depends on how well the user mimics the trained hand position Certain letters are more difficult to translate accurately because of their similarity Ex. E and S
Future Possibilites Translation into electronic devices such as computers and PDAs Wireless Implementation Expand from finger-spelling to whole words Arm motions Different syntax and grammar
Demonstration ASL Alphabet