Search

E-Learning is transforming the way education is imparted. Today, millions of students take self-paced online courses. However, the content and language complexity often hinders comprehension and this together with lack of immediate help from the course instructor leads to weak learning outcomes. Ability to predict difficult content in real time enables eLearning systems to adapt content as per students' level of learning. The recent introduction of low-cost eye trackers has opened the new class of applications based on eye response. Eye tracking devices can record eye response to the visual element or concept causing a learning difficulty. The response and the variations in eye response to the same concept over time may be indicative of the level of learning. In this paper, we use eye movement measures to predict the levels of learning associated with a term/concept. The main contribution of this study is the spatio-temporal analysis of eye response to a term/concept. Proposed system analyses slide images, extracts words (terms), maps the eye response to words, and prepares a term-response map. A majority voting classifier trained with terms of known learning levels uses this term response map to classify a term as novel or familiar. The proposed system achieves 61% accuracy when predicting learning difficulty.

We are surrounded by a very dense and detailed environment. However, when we perceive, we are unable to attend to all the details at once. Certain fully visible but unexpected objects get overlooked when we become involved in other tasks. Change Blindness and In-attentional blindness are psychological phenomenon where a visual stimulus goes unnoticed by the observer. In this paper, we present results of the effect of subliminal cueing at a targeted area of a video which is prone to go unnoticed due to in-attentional blindness. This helps us in understanding human visual processing response and has various applications such as surveillance and driving safety.

This is a MM'14 Workshop Summary Abstract for PIVP'14 - 1st International Workshop on Perception Inspired Video Processing. Workshop provided a venue for researchers involved with perceptual video processing and coding to present their current work and to discuss future directions. The workshop program consisted of a keynote talk, oral presentations of full papers and interactive poster session for the short papers. All participants took part in the discussion panel at the end of workshop, exchanging ideas and suggestions for future work.

In this paper, we present a subjective video quality evaluation system that has been integrated with different crowdsourcing platforms. We try to evaluate the feasibility of replacing the time consuming and expensive traditional tests with a faster and less expensive crowdsourcing alternative. CrowdFlower and Amazon's Mechanical Turk were used as the crowdsourcing platforms to collect data. The data was compared with the formal subjective tests conducted by MPEG as part of the video standardization process, as well as with previous results from a study we ran at the university level. High quality compressed videos with known Mean Opinion Score (MOS) are used as references instead of the original lossless videos in order to overcome intrinsic bandwidth limitations. The bitrates chosen for the experiment were selected targeting Internet use, since this is the environment in which users were going to be evaluating the videos. Evaluations showed that the results are consistent with formal subjective evaluation scores, and can be reproduced across different crowds with low variability, which makes this type of test setting very promising.

Media content distribution poses different challenges when it comes to supplying the variety of existing media formats requested by end-users. The need for an on demand transcoding stage may be justified under specific circumstances but is hampered by the high latency and long processing times. The great power cloud computing offers can be leveraged in a situation like this. This paper analyzes a use case where a Hadoop-based computer cloud is used as an aid in the transcoding of media content. The results point to the suitability of such a resource in a media distribution process requiring low cost streaming, such as the HTTP Live streaming scenario which supports streaming of audio or video to the iPhone, iPad, iPod touch and desktop computers as well as the transfer of video on demand with encryption and authentication.

Today, world is rapidly turning to high definition multimedia. From engineering and programming point of view, this usually means more computation is needed and more memory space is required to achieve these higher qualities. In this paper we explore the use of parallelization opportunities in graphics processors to accelerate video encoding. We evaluate the NVIDIA CUDA[1] toolkit and evaluate the performance of motion estimation in video encoding. The main goal of this paper is to evaluate the capabilities of NVIDIA/CUDA and develop a process for implementing video/multimedia applications. We have discovered that the difference in performance when CUDA is not used properly can be over 100x. We show how we were able to use CUDA capabilities to reduce the motion estimation time from 7000 milli seconds to 70 milli seconds.

Ambient Intelligence (AmI) systems refer to ubiquitous computing with focus on human experiences. The AmI ideas are based on the premise that if environment around individuals can be personalized for the current needs of an individual, then the lives can be more productive and enjoyable. New and existing technologies will be used to personalize spaces and experiences in AmI environments. The goal of AmI is to make computing and technology transparent and invisible to the individual. AmI is characterized by unobtrusive and mostly invisible technologies working together, anticipating the needs, and personalizing the environment and experiences for individuals. This paper presents new ideas in developing AmI infrastructure for deploying AmI services. The proposed system architecture is based on characterizing the roles and rights of individuals in the current environment and designing components and services that respond to these roles. The generalized architecture can be used to adapt physical as well as virtual spaces to fit individual preferences. A general architecture is necessary to move AmI from a collection of "cool gadgets" to an ambient and aware service focused on personalization for individuals.

Machine learning has been widely used in video analysis and search applications. In this paper, we describe a non-traditional use of machine learning in video processing - video encoding and transcoding. Video encoding and transcoding are computationally intensive processes and this complexity is increasing significantly with new compression standards such as H.264. Video encoders and transcoders have to manage the quality vs. complexity tradeoff carefully. Optimal encoding is prohibitively complex and sub-optimal coding decisions are usually used to reduce complexity but also sacrifices quality. Resource constrained devices cannot use all the advanced coding tools offered by the standards due to computational needs. We show that machine learning can be used to reduce the computational complexity of video coding and transcoding problems without significant loss in quality. We have developed the use of machine learning in video coding and transcoding and have evaluated it on several encoding and transcoding problems. We describe the general ideas in the application of machine learning and present the details of four different problems: 1) MPEG-2 to H.264 video transcoding, 2) H.263 to VP6 transcoding, 3) H.264 encoding and 4) Distributed Video Coding (DVC). Our results show that use of machine learning significantly reduces the complexity of encoders/transcoders and enables efficient video encoding on resource constrained devices such as mobile devices and video sensors.

This paper presents a novel macroblock mode decision algorithm for inter-frame prediction based on machine learning techniques to be used as part of a very low complexity MPEG-2 to H.264 video transcoder. Since coding mode decisions take up the most resources in video transcoding, a fast macro block (MB) mode estimation would lead to reduced complexity. The proposed approach is based on the hypothesis that MB coding mode decisions in H.264 video have a correlation with the distribution of the motion compensated residual in MPEG-2 video. We use machine learning tools to exploit the correlation and derive decision trees to classify the incoming MPEG-2 MBs into one of the 11 coding modes in H.264. The proposed approach reduces the H.264 MB mode computation process into a decision tree lookup with very low complexity. The proposed transcoder is compared with a reference transcoder comprised of a MPEG-2 decoder and an H.264 encoder. Our results show that the proposed transcoder reduces the H.264 encoding time by over 95% with negligible loss in quality and bitrate.

There is a renewed interest in the 3DTV research primarily due to the advances in low cost 3D display technologies. The two views required for 3DTV can be compressed using standard video compression techniques. MPEG-2 is widely used in digital TV applications today and H.264/MPEG-4 AVC is expected to be the leading video technology standard for digital video in the near future. The compression gains and quality of 3DTV will vary depending on the video coding standard used. While inter-view prediction will likely improve the compression efficiency, new approaches such as asymmetric view coding are necessary to greatly reduce bandwidth requirements for 3DTV. This demonstration will show the quality of 3D video experience on autostereoscopic displays using H.264 and MPEG-2. We will also show the benefits of using asymmetric view coding as well as the role of eye dominance on 3DTV experience.