Yi-Leh Wu

A comparison of DFT and DWT based similarity search in time-series databases

Similarity searc h in time-series databases has received signi cant attention lately .P opular tec hniques for e cien t retrieval of time sequences in time-series databases has been to use Discrete Fourier Transform (DFT). Recently, the Discrete Wavelet T ransform (DWT) has gained popular interest in database domain and several proposals have been made to replace DFT by DWT for similarity search over time-series databases. In this paper, we explore the feasibility of replacing DFT by DWT with a comprehensive analysis of the DFT and DWT as matching functions in time-series databases. Our results show that although the DWT based technique has several adv an tages,e.g., the D WThas complexity of O(N) whereas DFT is O(N logN), D WT does not reduce relativ e matching error and does not increase query precision in similarity searc h as suggested by previous works [1]. We conclude that, by exploring the conjugate property of DFT in real domain, the DFT-based and DWT-based techniques yield comparable results on similarity searc h in time-series databases.

PowerBookmarks: a system for personalizable Web information organization, sharing, and management

Users of the Web are overloaded with information. This medium is “polluted” with redundant, erroneous and low quality information. A WWW survey of 11,700 users conducted from April 10 to May 10, 1996[1] indicates that 30.31% of the users report “finding known info” is their problem and 27.80% of the users report organizing collected information as their problem. An empirical study[2] on users’ revisitation patterns to WWW pages found that 58% of an individual’s pages are revisits. With these study results, we believe the Web users would like to build and organize a larger collection of bookmarks for future references than they can reasonably maintain now.

Multi-camera spatio-temporal fusion and biased sequence-data learning for security surveillance

We present a framework for multi-camera video surveillance. The framework consists of three phases: detection, representation, and recognition. The detection phase handles multi-source spatio-temporal data fusion for efficiently and reliably extracting motion trajectories from video. The representation phase summarizes raw trajectory data to construct hierarchical, invariant, and content-rich descriptions of the motion events. Finally, the recognition phase deals with event classification and identification on the data descriptors. Because of space limits, we describe only briefly how we detect and represent events, but we provide in-depth treatment on the third phase: event recognition. For effective recognition, we devise a sequence-alignment kernel function to perform sequence data learning for identifying suspicious events. We show that when the positive training instances (i.e., suspicious events) are significantly outnumbered by the negative training instances (benign events), then SVMs (or any other learning methods) can suffer a high incidence of errors. To remedy this problem, we propose the kernel boundary alignment (KBA) algorithm to work with the sequence-alignment kernel. Through empirical study in a parking-lot surveillance setting, we show that our spatio-temporal fusion scheme and biased sequence-data learning method are highly effective in identifying suspicious events.

Adaptive learning of an accurate skin-color model

Due to variations of lighting conditions, camera hardware settings, and the range of skin coloration among human beings, a pre-defined skin-color model cannot accurately capture the wide distribution of skin colors in individual images. In this paper, we propose an adaptive skin-detection method, which allows modeling true skin-color distribution with significantly higher accuracy and flexibility than other methods attain. In principle, the proposed method follows a two-step process. For a given image, we first perform a rough skin classification using a generic skin model, which defines the skin-similar space. In the second step, a Gaussian mixture model (GMM), specific to the image under consideration and refined from the skin-similar space, is derived using the standard expectation-maximization (EM) algorithm. Then, we use an SVM (support vector machine) classifier to identify the skin Gaussian from the trained GMM (which contains two Gaussian components) by incorporating spatial and shape information of the skin pixels. This adaptive method can be applied to both still images and video applications. Results of extensive experiments performed on live video sequences and large image databases have demonstrated the effectiveness and benefits of the proposed model.

Discovery of a perceptual distance function for measuring image similarity

Abstract. For more than a decade, researchers have actively explored the area of image/video analysis and retrieval. Yet one fundamental problem remains largely unsolved: how to measure perceptual similarity between two objects. For this purpose, most researchers employ a Minkowski-type metric. Unfortunately, the Minkowski metric does not reliably find similarities in objects that are obviously alike. Through mining a large set of visual data, our team has discovered a perceptual distance function. We call the discovered function the dynamic partial function (DPF). When we empirically compare DPF to Minkowski-type distance functions in image retrieval and in video shot-transition detection using our image features, DPF performs significantly better. The effectiveness of DPF can be explained by similarity theories in cognitive psychology.

An adaptive skin model and its application to objectionable image filtering

We propose an adaptive skin-detection method, which allows modelling and detection of the true skin-color pixels with significantly higher accuracy and flexibility than previous methods. In principle, the proposed approach follows a two-step process. For a given image, we first perform a rough skin classification using a generic skin-model which defines the Skin-Similar space. The Skin-Similar space often contains many non-skin pixels due to the inevitable overlap in the color space between skin pixels and some non-skin pixels under the generic skin-model. The objective of the second step is to reduce the false-positive rate by analyzing the image under consideration. Specifically, in the second step, a Gaussian Mixture Model (GMM), specific to the image under consideration and refined from its Skin-Similar space, is derived using the standard Expectation-Maximization (EM) algorithm. We then use a Support Vector Machine (SVM) classifier to identify the skin Gaussian from the trained GMM by incorporating spatial and shape information of the skin pixels. Moreover, we examine how the improvement on skin detection by this adaptive skin-model impacts the detection accuracy in the application of Objectionable Image Filtering. We further propose a two-level classification scheme based on hierarchical bagging to improve the accuracy. Results of extensive experiments on large databases demonstrate the effectiveness and benefits of our adaptive skin-model.

Applying the golden rule of sampling for query estimation

Query size estimation is crucial for many database system components. In particular, query optimizers need efficient and accurate query size estimation when deciding among alternative query plans. In this paper we propose a novel sampling technique based on the golden rule of sampling, introduced by von Neumann in 1947, for estimating range queries. The proposed technique randomly samples the frequency domain using the cumulative frequency distribution and yields good estimates without any a priori knowledge of the actual underlying distribution of spatial objects. We show experimentally that the proposed sampling technique gives smaller approximation error than the Min-Skew histogram based and wavelet based approaches for both synthetic and real datasets. Moreover, the proposed technique can be easily extended for higher dimensional datasets.

Anatomy of a multicamera video surveillance system

Abstract.We present a framework for multicamera video surveillance. The framework consists of three phases: detection, representation, and recognition. The detection phase handles multisource spatiotemporal data fusion for efficiently and reliably extracting motion trajectories from video. The representation phase summarizes raw trajectory data to construct hierarchical, invariant, and content-rich descriptions of the motion events. Finally, the recognition phase deals with event classification and identification on the data descriptors. Through empirical study in a parking-lot-surveillance setting, we show that our spatiotemporal fusion scheme and biased sequence-data learning method are highly effective in identifying suspicious events.

Real-Time Hand Detection and Tracking against Complex Background

Most hand detection and tracking algorithms can be only applied in the fairly simple and similar background. We propose to combine a modified object detection method proposed by Viola and Jones with the skin-color detection method to perform hand detection and tracking against complex background. Out experimental results show that the proposed method is effective in near real-time speed (15 frames per second.).

Query estimation by adaptive sampling

The ability to provide accurate and efficient result estimations of user queries is very important for the query optimizer in database systems. In this paper, we show that the traditional estimation techniques with data reduction points of view do not produce satisfiable estimation results if the query patterns are dynamically changing. We further show that to reduce query estimation error, instead of accurately capturing the data distribution, it is more effective to capture the user query patterns. In this paper, we propose query estimation techniques that can adapt to user query patterns for more accurate estimates of the size of selection or range queries over databases.