Recognition of Fingerspellings in Japanese Sign Language based on the Detection of Nail and Wrist Position

Abstract

We have recognized fingerspellings in Japanese sign language from hand images. We can discriminate the front and back of hand by detecting nails. Also, by detecting the positions of the wrist and fingertips, we generate vectors for the recognition. Then, the vector generated based on the nail and wrist positions is compared with the one in database, and then the fingerspelling is recognized. As the result of the experiment for 41 fingerspellings without movement, the recognition rate was 75.6% for 7 examinees.