We propose a novel technique for head pose classification in crowded public space under poor lighting and in low-resolution video images. Unlike previous approaches, we avoid the need for explicit segmentation of skin and hair regions from a head image and implicitly encode spatial information using a grid map for more robustness given lowresolution images. Specifically, a new head pose descriptor is formulated using similarity distance maps by indexing each pixel of a head image to the mean appearance templates of head images at different poses. These distance feature maps are then used to train a multi-class Support Vector Machine for pose classification. Our approach is evaluated against established techniques [3, 13, 14] using the i-LIDS underground scene dataset [9] under challenging lighting and viewing conditions. The results demonstrate that our model gives significant improvement in head pose estimation accuracy, with over 80% pose recognition rate against 32% from the best of existing models.

In this paper, we address the recognition of subtle facial expressions by reasoning on the classification confidence. Psychological evidences have determined that eyelids and eyebrows are significant for the recognition of subtle facial expressions and the early perception of human emotions. This early perception results in a more complex problem, which requires a confidence assessment for any provided solution. Thus, traditional score-based classifiers (e.g. k-NN and NN) are not able to produce confident estimates. Instead, we first present five confidence estimators and a confidence classification assessment for Case-Based Reasoning (CBR). Second, we improve the expression retrieval from the database by learning the neighbourhood's dimensions for the expected classification confidences. Third, we reuse the previous classified expressions and the confidence assessment to improve the classification achieved by k-NN. Fourth, we improve the database for generalization with new subjects by learning thresholds to minimize misclassification with low confidence, maximize correct classifications with high confidence and re-arrange misclassification with high confidence. The proposed system represents an effective contribution for both subtle expression recognition and CBR methodology. It achieves an average recognition of 97% plusmn 1% with a confidence of 96% plusmn 2% for expressiveness between 20% and 100%.

Face detection and tracking, through image sequences, are primary steps in many applications such as video surveillance, human computer interface, and expression analysis. Many currently existing techniques donpsilat perform well due to pose variations, appearance changes, illumination changes, complex backgrounds, and inaccurate initialization. The last short coming, which is the difficulty to initialize motion regions, is a problem facing any tracker. In this paper, we present an automatic and robust face detection and tracking system for color image sequences. Face detection is done using skin color segmentation and connected components analysis. Later, facial features are detected by active shape models and a face mesh is initialized. Finally, the tracking is done by active appearance models. Experimental detection and tracking results on a pose varying face video are given.

The paper addresses the problem of on-line scheduling of mandatory/reward-based systems in which tasks have a hard real-time mandatory part and an optional part with a non-decreasing reward function associated to its execution. Four methods, generically called SH, are proposed. They are based on the detection of singularities, special instants that appear along the execution of the system. By applying some heuristic rules, the singularities and some of the following slots are used to process optional parts. To the best of the authors' knowledge, papers published up to now on the subject of reward maximization require the functions to be continuously differentiable. Moreover, they cannot be executed on-line. On the contrary, the only requirement of the SH methods is that the functions are computable at every instant. After the system is proved to be schedulable by any of the available exact test off-line techniques, SH methods can be executed on-line. Their performance is evaluated using simulations performed on a synthetic set of tasks proposed in one of the outstanding papers on the subject and on sets of tasks randomly generated. The results are analysed and explained. The main conclusionis that , in all cases, the SH methods outperform the Best Incremental Return, often used as a yardstick.

This paper presents a probabilistic model of a client/server multimedia system based on the theory of Markov processes. The server provides different qualities of service (hard real-time, soft real-time and best-effort) and is operated using a new scheduling concept based on a hybrid deterministic/probabilistic analysis. The model also provides answers to some fundamental engineering design questions related to disk price/performance tradeoffs.

When applying the Rate Monotonic discipline to schedule a set of periodic preemptible real-time tasks, the scheduler may be able to distinguish only a limited number of priority levels. This is common in control applications using low cost embedded controllers. If the number of tasks to be scheduled is larger than the number of distinguishable levels, the set of tasks must be partitioned in a set of priority classes. RM can be used only to arbitrate conflicts between tasks of different classes. In this paper a method to determine the minimum number of priority levels necessary to schedule the set of tasks is formally proved and its complexity analysed. Finally, a systematic method to obtain all the possible partitions with the minimum number of classes, resembling the Quine's method to minimize Boolean functions, is also given.

The purpose of the paper is to present a method to validate tentative routes to establish calls with end-to-end timing-constraints in a multihop packet-switching network. Timing release counters are proposed as a mechanism to guarantee that intervals between consecutive releases of messages of the same call are not shorter than the call period. A comparative evaluation of the proposed timing release counters vs. the alternative method of stop-and-go queues is performed.

Hard real-time multitask-multiprocessor systems are one of the more important species of the taxonomic genus of distributed systems. The assignment of m preemptible periodic tasks to be executed in a set of n processors (interconnected through a communications network) is a NP-complete problem that must meet several constraints: time, allocation, resources, communication and precedence. The first three are processor-dependent and the fourth is network-dependent, but all of them can be verified as the assignment process goes on. Precedence constraints, instead, can be best verified when a tentative assignment, meeting the first four constraints, is obtained. In this paper, a formally proved method is presented to validate tentative assignments. In order to solve the problem of jitter in data arrivals, timed release counters are proposed.

A comparative evaluation of standard local area networks operating in time-constrained environments is performed. The analyzed protocols are token bus (IEEE 802.4), token ring (IEEE 802.5), and a modified version of CSMA/CD (carrier-sense multiple-access with collision detection) (IEEE 802.3). The priority disciplines necessary to make the systems crisis-free are discussed in connection with the relationships between the slot time, the number of nodes in the network, and their crisis times. In each case, the efficiency and the production of the system (in Mb/s) as functions of the length of the data field are determined.>