Research Projects
Human Computer Interaction State-of-the-art Research
DIU HCI RL is working in a tremendous speed to gather the state-of-the-art study on this area. This project covers different interesting and challenging topics that are recently gaining popularity among the researchers all over the world. Recent research problems are justified under the experimental process of different methodologies and approaches proposed by researchers all over the world.
Researchers: Undergraduate Students
Mentor: Dr. Sheak Rashed Haider Noori
Human Activity Recognition using Mobile Phone and Inertial Sensors
Human activity recognition (HAR) is a challenging problem to solve in modern days. Humans are involved in everyday works which are nothing but different actions performed in sequences. Understanding those sequence of actions to understand human activity using machine is not an easy task. Identifying different steps and recognizing the activity using regular mobile phone sensors like accelerometer, gyroscope and magnetometer is important as people nowadays uses mobile phone more than any other time. In this project we are trying to intrigate the sensor values coming from the mobile phones built-in sensors and extra inertial sensors. Applying machine learning approaches are the main way-out here.
Researchers: Md. Ferdousur Rahman Sarker, Shafayet Hossain
Mentor: Ahmed Al Marouf
Cricket shot classification using 3D Depth Imaging with Microsoft Kinect and Deep Learning
Cricket, one of the most popular games, the game of bat and ball, the game of strategy. Application of computer vision and machine learning techniques in cricket for different analysis is an emerging domain now. Cricket shot detection from single video is an important aspect, but not much work has been done in this domain yet. It is hard to classify cricket shot as occultation of bat as well as the different types of body. We though can do skeleton based action detection as well as calculate the bat angle in between. Therefore, we are proposing a novel approach to classify cricket shots using 3D depth imaging with Microsoft Kinect and Deep Learning.
Researchers: Mohammad Shakirul Islam (151-15-311), Md. Ferdouse Ahmed Foysal (151-15-5274)
Mentor: Nafis Neehal (NN)
Assistive Framework for Hearing Impaired: A Deep Learning Based Approach for Detecting Bengali Vowels using Visual Data from Lip Reading Technique with MS Kinect
Lip-Reading (LR) is a technique which is related to human computer interaction (HCI). Lip reading can help the hearing impaired to understand the speech without hearing it. A lot of research has been done on lip-reading in English and some on Turkish, French, Chinese and few other languages, but not much research has been done on lip-reading in Bengali. So, our aim is to develop an application for detecting Bengali Vowels using visual data from Lip-Reading (LR) technique with MS Kinect. Lip reading of Bengali Vowels is challenging because of their small range of variations in the pronunciation as they sound almost similar. Deep Learning is a new area of Machine Learning research and TensorFlow is a framework created by Google for creating Deep Learning models, currently the most popular deep learning framework for computer vision. In this work we provide a Deep Learning based novel approach for detecting Bengali Vowels using Lip Reading technique with MS Kinect to extract Visual Data and analyze it with TensorFlow framework.
Researchers: Shahana Shultana (151-15-5155), Md. Shakil Moharram (142-15-3662)
Mentor: Nafis Neehal (NN)