The thesis introduces a novel search paradigm, and explores a previously unreported behavioural error detectable in this paradigm. In particular, the ‘Unpacking Task’ is introduced – a search task in which participants use a computer mouse to sort through random heaps of items in order to locate a unique target. The task differs from traditional search paradigms by including an active motor component in addition to purely perceptual inspection. While completing this task, participants are often found to select and move the unique target item without recognizing it, at times continuing to make many additional moves before correcting the error. This ‘unpacking error’ is explored with perceptual, memory load, and instructional manipulations, evaluating eye-movements and motor characteristics in additional to traditional response time and error rate metrics. It is concluded that the unpacking error arises because perceptual and motor systems fail to adequately coordinate during completion of the task. In particular, the motor system is found to ‘process’ items (i.e., to select and discard them) more quickly than the perceptual system is able to reliably identify those same items. On those occasions where the motor system selects and rejects the target item before the perceptual system has had time to resolve its identity, the unpacking error results. These findings have important implications for naturalistic search, where motor interaction is common, and provide further insights into the conditions under which perceptual and motor systems will interact in a coordinated or an uncoordinated fashion.

<p>In computer vision, the pictorial structure model represents an object in an image by parts that are arranged in a deformable configuration. Each part describes an object's local photometric appearance, and the configuration encodes the global geometric layout. This model has been very successful in recent object recognition systems.</p><p>We extend the pictorial structure model in three aspects. First, when the model contains only a single part, we develop new methods ranging from regularized subwindow search, nested window search, to twisted window search, for handling richer priors and more flexible shapes. Second, we develop the notion of a weak pictorial structure, as opposed to the strong one, for the characterization of a loose geometric layout in a rotationally invariant way. Third, we develop nested models to encode topological inclusion relations between parts to represent richer patterns.</p><p>We show that all the extended models can be efficiently matched to images by using dynamic programming and variants of the generalized distance transform, which computes the lower envelope of transformed cones on a dense image grid. This transform turns out to be important for a wide variety of computer vision tasks and often accelerates the computation at hand by an order of magnitude. We demonstrate improved results in either quality or speed, and sometimes both, in object matching, saliency measure, online and offline tracking, object localization and recognition.</p> / Dissertation

Nowadays, search engines are more relying on machine learning techniques to construct a model, using past user queries and clicks as training data, for ranking web pages. There are several learning to rank methods for information retrieval, and among them ranking support vector machine (SVM) attracts a lot of attention in the information retrieval community. One difficulty with Ranking SVM is that the computation cost is very high for constructing a ranking model due to the huge number of training data pairs when the size of training dataset is large. We adopt the MapReduce programming model to solve this difficulty. MapReduce is a distributed computing framework introduced by Google and is commonly adopted in cloud computing centers. It can deal easily with large-scale datasets using a large number of computers. Moreover, it hides the messy details of parallelization, fault-tolerance, data distribution, and load balancing from the programmer and allows him/her to focus on only the underlying problem to be solved. In this paper, we apply MapReduce to Ranking SVM for processing large-scale datasets. We specify the Map function to solve the dual sub problems involved in Ranking SVM and the Reduce function to aggregate all the outputs having the same intermediate key from Map functions of distributed machines. Experimental results show efficiency improvement on ranking SVM by our proposed approach.

Meaning based search retrieves objects from search index repository based on user's search Meanings and meaning of objects rather than keyword matching. It requires techniques to capture user's search Meanings and meanings of objects, transform them to a representation that can be stored and compared efficiently on computers. Meaning of objects can be adequately captured in terms of a hierarchical composition structure called concept tree. This thesis describes the design and development of an algorithm that transforms the hierarchical concept tree to a tensor representation using tensor algebra theory. These tensor representations can capture the information need of a user in a better way and can be used for similarity comparisons in meaning based search. A preliminary evaluation showed that the proposed framework outperforms the TF-IDF vector model in 95% of the cases and vector based conceptual search model in 92% of the cases in adequately comparing meaning of objects. The tensor conversion tool also was used to verify the salient properties of the meaning comparison framework. The results show that the salient properties are consistent with the tensor similarity values of the meaning comparison framework.

In mobile communications networks, a location management scheme is responsible for tracking mobile users. Typically, a location management scheme consists of a location update scheme and a paging scheme. Gau and Haas first proposed the concurrent search(CS) approach that could simultaneously locate a number of mobile users in mobile communications networks. We propose to use the theory of the discrete-time Markov chain to analyze the performance of the concurrent search approach. In particular, we concentrate on the worst case in which each mobile user appears equally likely in all the cells of the network. We analyze the average paging delay, the call blocking probability and the system size. We show that our analytical results are consistent with the simulation results of the concurrent search.

Digital image data requires large compression ratio in applications like internet, communication and audio-visual environment. In this thesis, we realize the MPEG-4 codec standard on the ARM9-based platform and improve the execution performance by efficient implementations of the core operations such as Motion Estimation and DCT. In the assembly codes obtained by directly compiling the C codes, there exists a lot of redundant checking which causes a large amount of execution time waste. We rewrite some of the compiled assembly codes to improve the execution efficiency using a variety of techniques such as loop-unrolling and data-type optimization. We also analyze the experimental results using several benchmark video sequences with different modes.

NONE

The performance and speed of the interframe motion estimation method for sequencial frame sequence compression are the important issues especially in networking application such as video conference and video on demand. In this paper, we proposed a new fast search algorithm for motion estimation on block matching technique called the cocktail search algorithm (CTS). This new algorithm takes advantages of prior search algorithms proposed in the literature and improves at our observations. The experiment results show that the proposed CTS algorithm can provide the better performance and require less computational costs than others. In other words, the CTS can obtain the accurate motion vector efficiently and fast. The fruitful results is achieved by not only holding the original benefit but also constructively improving the existing drawbacks.

none

Internet has become the most extensible media of data exchange and communication in the world because computer science and technology are more and more popular. The Geographic Information Systems¡]GIS¡^are also developed on the Internet. However, using existing mechanism of data searching on the Internet cannot search data in Web GIS. We can only browse data but not access. This situation makes Web GIS as an isolated island. Users fail to know where and what kinds of data are provided, and these data also cannot be shared. The most important objective of the research is to build an effective mechanism of searching and clearinghouse on the Internet. This mechanism can help computer overcome difficulties in reading and understanding geographic data that are composed of maps and images, and then geographic data can be searched and shared easily as text data. The research will try to create metadata by XML that are complied with FGDC standard. By using two of the XML characteristics, i.e. creating tags and describing data, the computer can retrieve information automatically from metadata on the Internet. Lastly, the geographic search engine and clearinghouse that the research built will collect and integrate geographic metadata to systematically facilitate users finding geographic data they need through Internet, and achieve the objectives of geographic data search and clearinghouse on the Internet.