Pradeep Kumar

I worked as a Postdoc researcher in the Institute of Biomedical Engineering in University of New Brunswick with Dr. Erik Scheme. My past research projects were gesture recognition, Multiple person activity recognition and 3D handwriting recognition systems with the help of EMG, Microsoft Kinect and Leap motion sensors. To model the data, we used a family of sequential classification techniques including Hidden Markov Models (HMM), Recurrent Neural Networks (RNNs) (e.g. LSTMs, BLSTMs) and Hidden-State Conditional Random Fields (HCRFs).

In addition to this, I carried my research in Cognitive Engineering with EEG brain sensor for the development of various Brain-Computer-Interface (BCI) applications including Biometrics, Prediction analysis and Security, etc. 

I did my Ph.D. in Computer Science & Engineering (CSE) at Indian Institute of Technology, Roorkee (IIT Roorkee) under the supervision of Dr. Partha Pratim Roy. During my research, I worked in CVPR Lab, ISI Kolkata with Prof. Umapada Pal and in CRCV Lab at University of Central Florida with Prof. Mubarak Shah. My research area include Pattern Recognition.

Demonstration: 

1. In this demonstration, we are showing the recognition of Sign Language gestures using Leap motion sensor. The recognition has been performed with the help of Hidden Markov Model (HMM) for isolated sign words. The sign gestures are performed by a signer in Indian Sign Language and the corresponding recognition appear on the monitor screen in the form of text.

2. In this demo, we are showing the recognition of finger spelled word gestures (in English) in air with the help of Leap motion sensor. The recognition process has been carried out using HMM classifier.

3. A 3D Avatar has been developed for the learning purpose of sign language. The application has easy to use interface where a user can enter a sentence or any isolated word to know its corresponding sign language representation.