Technical Papers Parallel Session-I: Arm gesture recognition on microsoft KinectUsinga Hidden Markov Model-based representations of poses
Abstract/Description
Gesture recognition has recently generated significant research interest. It is primarily apprehensive on exploring the performance of human acumen. Visual understanding of hand gestures can help in attaining the simplicity and characteristic craved for Human Computer Interaction (HCI). Gesture recognition is effortless for human beings but a very challenging task when it comes to computers. To aid this problem, we have proposed a Kinect based state-of-the-art solution. We introduce three gestures i.e. acceleration, turn right and turn left and yield their skeletal tracks through Kinect. Collected dataset is then normalized and trained to accumulate library of poses using an HMM-based algorithm. We evaluate our approach on a dataset of 228 videos. After cross-validation, experimental results show that the accuracy of 81.13% is achieved for discretized poses.
Keywords
Hidden Markov models, Gesture recognition, Videos, Computational modeling, Libraries, Indexes, Acceleration
Location
C-9, AMAN CED
Session Theme
Technical Papers Parallel Session-I (Artificial Intelligence)
Session Type
Parallel Technical Session
Session Chair
Dr. Jawwad Shamsi
Start Date
12-12-2015 2:30 PM
End Date
12-12-2015 3:50 PM
Recommended Citation
Siddiqui, S. A., Snober, Y., Raza, S., Khan, F. M., & Syed, T. Q. (2015). Technical Papers Parallel Session-I: Arm gesture recognition on microsoft KinectUsinga Hidden Markov Model-based representations of poses. International Conference on Information and Communication Technologies. Retrieved from https://ir.iba.edu.pk/icict/2015/2015/3
COinS
Technical Papers Parallel Session-I: Arm gesture recognition on microsoft KinectUsinga Hidden Markov Model-based representations of poses
C-9, AMAN CED
Gesture recognition has recently generated significant research interest. It is primarily apprehensive on exploring the performance of human acumen. Visual understanding of hand gestures can help in attaining the simplicity and characteristic craved for Human Computer Interaction (HCI). Gesture recognition is effortless for human beings but a very challenging task when it comes to computers. To aid this problem, we have proposed a Kinect based state-of-the-art solution. We introduce three gestures i.e. acceleration, turn right and turn left and yield their skeletal tracks through Kinect. Collected dataset is then normalized and trained to accumulate library of poses using an HMM-based algorithm. We evaluate our approach on a dataset of 228 videos. After cross-validation, experimental results show that the accuracy of 81.13% is achieved for discretized poses.
