Indian Sign Language Recognition Using Kinect Sensor

There has been a lot of research on automatic recognition of Sign languages and is an effective means of transferring information for Deaf and Hard of Hearing (HoH) community. Here we propose a system for Indian Sign Language recognition, which uses Microsoft Kinect sensor and Machine learning for effectively recognizing some signs used in Indian Sign Language. Kinect generates the skeleton of a human body and detects 20 joints in it. We use 11 out of 20 joints and extract 34 novel features per frame, based on distances and angles involving upper body joints. These features are trained with a multi-class Support Vector Machine achieving an accuracy of 100% and 86.16% on train and test data respectively. Proposed system recognizes 37 signs in real time. The data is used in the proposed system is generated by the Deaf and Hard of Hearing (HoH) persons in our lab.