Balaji Hariharan

Gesture recognition using Kinect in a virtual classroom environment

In an E-learning classroom, students and professors are at different geographic locations. In the present E-learning system, the professor and the student interact with each other through simple internet applications. The disadvantage in this type of classroom set-up is that students in a remote location will not receive the same attention as those ones in a real classroom. Students in a remote location will only be seen on a small screen and the clarity of the video might be poor. Furthermore, the professor in a real classroom might focus only on students who are interacting face-to-face with him. The professor may not know the details of the remote students and how well they understand his class lecture. To solve this problem, this paper describes a system, based on KinectTM, to make the E-learning class lecture more interactive for the remote students. For example, when a student in a remote location, raises his/her hand to talk to a professor, the Kinect camera detects the student, identifies the gesture of the remote student, zooms onto the remote student, and displays details such as his/her name and background of study. Thus the professor can pay equal attention to both the remote and local students.

Wireless power transfer to underground sensors using resonant magnetic induction

Wireless communication systems play an increasingly integral role in society. Currently underground sensor networks are used in many applications for security purposes, environmental applications, infrastructure maintenance etc. The major problem, associated with sensor networks, is in providing the power supply to enable the networks to work for long durations. This paper presents a novel technique to wirelessly, supply power to the underground sensor networks using the magnetic induction principle. In addition, using the resonating principle further ensures maximum efficiency of this proposed system. Experimental results to prove the feasibility of this novel technique for supplying wireless power are presented.

Delay and energy optimization in multilevel balanced WSNs for landslide monitoring

In most of the real world wireless sensor network deployments, the energy utilization is a critical factor as the nodes are battery powered. In most of the real-world deployments it is observed that the sensing subsystem consumes higher power. In order to extend the lifetime of such systems it is required to reduce the sensing energy than communication energy. We have deployed a system for monitoring Landslides in India consists of 150 geo-physical sensors and used solar panels to power these sensor nodes. The decision making in favor of Landslide occurrence is based on the maximum values obtained from the high priority sensors. As this maximum value is not frequently changing in the deployment, locating the sensor node with maximum value allows us to switch off the other sensors for a predetermined period of time. This work proposes an optimal balanced network topology for delay minimization by parallelizing data aggregation operation in each sub-network. The sensor node switch off schemes on the top of delay minimized topology enables the optimal utilization of the available solar power. The analysis of these mechanisms shows that, more number of nodes can be powered with the available source of energy and can increase the network life time

Efficient wireless power transfer using underground relay coils

Underground Wireless Sensor Network (UWSN) is a newly emerging technology that is capable of replacing existing traditional wired connections. Research has proved that magnetic induction based communication performs better than electromagnetic wave communication, especially in the dynamic underground environment. Currently, data recorders are being used to collect the data from underground sensors. Data recorders have a few disadvantages such as the inability to produce data on time and the difficulty to deploy them. Many problems related to data recorders can be solved using UWSNs magnetic induction technique. One of the main challenges associated with the UWSN is the lack of efficient power transfer to underground sensors as the distance between the transmitter and the receiver coils increases. This paper introduces a unique technique where relay resonators are used between the transmitter and the receiver coils to increase the efficiency for charging sensors which are present underground. Experimental results that prove the presence of relay coils for efficient power transfer are presented.

Remote student localization using audio and video processing for synchronous interactive e-learning

The advancements in the field of wireless technology has led to a widespread acceptance of multimedia on wireless networks. The increasing popularity of multimedia networking applications such as e-learning has brought in more requirements in educational institutions for the students to learn online Current e-learning systems do not facilitate synchronous learning i.e., there is no interaction like that of a real class. However to impart effective learning, interaction between a student in the remote class-room and the teacher in the main classroom is mandatory. To enhance this virtual classroom, an automatic e-learning system is been proposed. Wireless technology which consists of collection of audio and video sensors tracks and identifies the speaker of interest in the classroom so that the lecturer on the other end will be able to effectively interact and can have a one to one communication with the focused student. We describe in this paper an automated system for student localization using audio and video processing, as a first step towards achieving such an automated system.

Scalable, Secure, Fail Safe, and High Performance Architecture for Storage, Analysis, and Alerts in a Multi-site Landslide Monitoring System

Open image in new window Wireless sensor networks can be deployed in landslide prone areas to monitor various geological and weather properties to detect a possible landslide and provide early warning to local public for evacuation. Implementing and managing a system for capturing sensor data from the deployment sites and transferring to the central database for storage, analysis, retrieval and prediction is an endeavor riddled with both natural and technical challenges. The usual remoteness of the landslide prone areas result in power restrictions, bandwidth constraints and frequent connectivity issues. As the sensors and systems are exposed to constant natural elements, they are prone to frequent failures and calibration issues. Our high performance, scalable, robust, and secure system; featuring multi-site landslide data capture, replication, storage, monitoring, and processing functionalities, surmounts all these challenges effectively. The scalability and performance is achieved by real-time streaming of compressed data, in-memory processing, bulk storage, and retrieval through partitioned tables. The security is achieved through authenticating and encrypting streamed data and keeping only minimal raw data on site. The fail-safety is achieved through automated reconnection, and persisting and cross-tracking data at each processing step. Finally the high performance in analysis and alerts are achieved by series of hierarchical and temporal aggregate tables. In this paper we present the architecture and features of our landslide data system along with the performance testing statistics and related analysis. We demonstrate that our system is capable of handling data from 100 sites, each having 1000 sensors and sending data once a minute using a single cloud server.

Re-orchestration of Remote Teaching Environment in eLearning

eLearning has become indispensable in making lectures from experts accessible to a wide student audience. Current eLearning solutions offer many ways to re-create the teaching environment in a remote location. They focus on recreating the elements that make up the teaching side. In this paper we identify the components of a teaching environment and the factors that constitute a conducive learning environment for students. We identify the common methods of teaching side reconstruction and conduct a study to evaluate the effect of each on the learning outcome on these factors. We analyse the network and computational performance of these methods and arrive at a most optimal method to use in a eLearning session.

Multimedia: Video Sharing for E-classroom

Nowadays, e-learning has become one of the common platforms that every university relies on for long distance, and real time education transfer. This paper solves one of the important research challenges in this area, which is the loss of synchronization between the videos shared among different classrooms and the avoidance of network throttling. To achieve an interactive e-learning environment, a Real Time Media Sharing Protocol (RTMS Protocol) is proposed in this paper. This RTMS Protocol facilitates the stable and scalable synchronization of shared video and video operations, from a real classroom to multiple virtual classrooms. The paper contains a sequence diagram and timing diagram of the RTMS Protocol. The protocol is implemented in Adobe Flex and is tested in a real world scenario with one real classroom and two virtual classrooms. The use of Round Trip Time (RTT) and Universal Resource Locators (URL) advance the design of this protocol.

Resolution Scaled Quality Adaptation for Ensuring Video Availability in Real-time Systems

Online videos are one of the most prevalent forms of communication over the Internet because humans are attracted to visual stimulations. Multimedia services, especially real-time services, require data to arrive on time. Unlike any other transmissions, video streaming needs to be monitored and adapted to bandwidth, when the video stream passes through the best-effort environment. A guaranteed QoS (Quality of Service) is essential in all real-time applications. This paper presents a context aware mechanism, which improves the QoS of real time video transmission, over wireless networks. This proposed RSQA system ensures video availability even when the bandwidth is low. Drawbacks to other solutions for saving bandwidth in real-time include being: (1) too expensive due to requiring a priority based router (in the case of proxy based and frame prioritization approaches); (2) too complex; (3) not appropriate for real-time systems. This RSQA system succeeds in being of low complexity, appropriate for real-time systems and cost effective (not requiring a priority based router to function). This RSQA system is independent of wireless technology making, thus it appropriate for use in a heterogeneous environment as well. Whenever a user experiences packet loss or a delay in receiving a video, the server will be notified about the congestion and decrease the video quality to make the video available at low bandwidth. The decision regarding the needed QoS parameters and the rate control is based on context information (available bandwidth). The RTP (Real-time Transport Protocol) based streaming is simulated and verified by an Eclipse simulator and Java using an MJPEG standard. Results show that the QoS parameters of video transmission were improved considerably by this RSQA system.

WiEyeTNB: A Wireless Sensor Based Drowning Detection System for Enhanced Parental Care

Drowning is considered as the third leading cause of unintentional injury death globally. In many low and middle income countries drowning is one of the leading causes of death, especially for children under 12 years old. According to the World Health Organization (WHO) report 388000 people died in 2004 as a result of drowning around the world. The overall global rate of drowning among children is 7.8 deaths per 10000 populations. Advances in sensing devices and integrated circuit technology pave the way for many smart sensing products. WiEyeTNB, Wireless Eye That Never Blinks, discusses the novel notion of designing a smart sensor product for detecting drowning in advance, especially among the children under the age of 5. Smart sensor devices consist of a microprocessor, RF transceiver, one or more sensors and powering devices. The proposed system issues an alarm through a wireless network so that immediate attention can be given to the casualty. Simulation of the system is performed using QualNet 5.0.2 Network Simulator and evaluated the performance of the proposed system architecture.