German Finger Spelling Recognition System (GFRS)

Screenshot from 2015-10-30 17:42:14

This project consisted into evaluating the performances of the Leap Motion Controller in the context of isolated and continuous handshape recognition. An HMM-based letter to letter transition model was used in order to describe the dynamics of the hand motion.

Experiments conducted on both isolated and continuous recognition. For isolated recognition, the system could achieve an accuracy of 80% using a vocabulary of 100 transitions and the accuracy could be further improved to 89.96% when reducing the vocabulary size to 30. When applied on continuous recognition, the accuracy of the system fell to 68%.

Tengfei’s (王腾飞) master thesis is now available for download (link below). The presentation will be held at DFKI in Room Turing II on November the 11th (2015) at 14:00.

Hidden Markov Model Based Recognition of German Finger Spelling Using the Leap Motion

Advertisements
This entry was posted in Research. Bookmark the permalink.