For many applications, it is important to quickly locate the nearest neighbor of a given time series. When the given time series is a streaming one, nearest neighbors may need to be found continuously at all time positions. Such a standing request is called a continuous nearest neighbor query. This paper seeks fast evaluation of continuous queries on large databases. The initial strategy is to use the result of one evaluation to restrict the search space for the next. A more fundamental idea is to extend the existing indexing methods, used in many traditional nearest neighbor algorithms, with pre-fetching. Specifically, pre-fetching is to predict the next value of the stream before it arrives, and to process the query as if the predicted value were the real one in order to load the needed index pages and time series into the allocated cache memory. Furthermore, if the pre-fetched candidates cannot fit into the cache memory, they are stored in a sequential file to facilitate fast access to them. Experiments show that prefetching improves the response time greatly over the direct use of traditional algorithms, even if the caching provided by the operating system is taken into consideration.

No abstract

No abstract

In this paper we consider the problem of tracking of moving human face in front of a video camera in real-time for a Model-based coding (MBC) application. The 3D head tracking in a MBC system could be implemented sequentially as 2D location tracking, coarse 3D orientation estimation and accurate 3D motion estimation. This work focuses on the 2D location tracking of one subject face object through continuously using a face detector. The face detection scheme is based on a boosted cascade of simple Haar-like feature classifiers. Although such detector demonstrated rapid processing speed, high detection rate can only be achieved for rather strictly near front faces. This introduces the "loss of tracking" problem in 2D tracking when the face rotate a big angle. This paper suggests an easy accessory solution to overcomes the pose problem by using Dynamic Programming (DP). The Haar-like facial features are spatially arranging into a 1D deformable face graph and the DP matching is used to handle the "loss of track" problem. DP match the deformed version of the face graph extracted from a rotated face with the reference one took online when "loss of tracking" happen. Since the deformable face graph covers big pose variation, the developed technique is robust in tracking rotated faces. Embedding Haar-like facial features into a deformable face graph is the key feature of our tracking scheme. A real time tracking system based on this technique has been set up and tested. Encouraging results have been got and are reported.

No abstract