Charles Z. Liu

Extensions of principle component analysis with applications on vision based computing

This paper mainly focuses on the principle component analysis (PCA) and its applications on vision based computing. The underlying mechanism of PCA given and several significant factors, involved with subspace training are discussed theoretically in detail including principle components energy, residuals assessment, and decomposition computation. The typical extensions, including probabilistic PCA (PPCA), kernel PCA (KPCA), multi-dimensional PCA and robust PCA (RPCA), have been presented with critical analysis on both mechanisms and computations. Combining with the studies on, such as, image compression, visual tracking, image recognition and super-resolution image reconstruction, PCA and its extensions applied to computer vision are critically reviewed and evaluated on the targeted issues in each application as well as the role they played at specific tasks to the characteristics, cost and suitable situations of each PCA based vision processing technique.

Data-aware QoE-QoS management

In this paper, we propose a data aware method to perform Quality of Experience (QoE) and/or Quality of Service (QoS) management. A data-aware QoE model is built based on Principle Component Analysis (PCA). Three main types of QoE-QoS management (optimal QoE management, optimal QoS management and QoE-QoS balance management) are discussed based on the proposed scheme. A QoE-QoS unified optimization is also presented in normalized scale space. An experiment is given to show how the proposed scheme works. With the proposed scheme, the operations for QoE-QoS management is simplified but flexible, and the optimal solution can be obtained analytically with balanced efficiency.

Scalable Learning for Dispersed Knowledge Systems

This paper mainly focuses on dealing with the issue of scalable learning in dispersed knowledge system. A scalable learning scheme and ξ process are proposed with a theoretical analysis. With the proposed scheme, the dispersed knowledge system can be used as a centralized system without knowing the overview of the global database. A case study of application in dispersed face recognition system is given to show how the proposed scheme implements and works.

Motion-keying Based Dynamical Scene Layering with Adaptive Learning

Chroma-keying has been widely applied in multimedia. However, its constraints, including stadium setup, mono-chroma detection and specific techniques for video capture, limit the utilization in consumer electronic applications. This paper presents an economical scheme to perform the keying effect based on motion tracking with low requirements on the conditions for keying. An adaptive learning strategy is also developed to perform a layering task without the support of prior knowledge of the scene. An example is given to demonstrate the effect.

A data-aware confidential tunnel for wireless sensor media networks

A Wireless Sensor Media Network (WSMN) provides an economical and feasible solution to microscopic visual monitoring in real time. Compared to macrographic monitoring (e.g. aerial photography), it can also provide a flexible solution by means of the facility of setup and integration of microscopic data. Security is one of the significant issues for the applications of monitoring due to the visuality of the content in a WSMN. Besides the visual data, the confidentiality of the communication is also important since a WSMN serves as both a media network and a sensor network. Impersonation and key recovery attack are two particular harmful threats to the sensor networks. A spying agent masquerades as a normal node in the network, overhearing the data and signaling. In this paper, we present a data-aware scheme for a wireless sensor media network (WSMN) to guarantee the security of information system against both the impersonation and key recovery attacks during communication. With the consideration of the characteristics of visual data regarding both volatility and correlation, a chaos-driven strategy is used to design the protocol for tunnel establishment. The data-aware scheme is proposed to enhance the security for confidential communication by public key cryptography based attack detection. A simulation and related analysis are presented in this paper. The robustness against anti-attack is also discussed. The results show that the scheme is able to protect the system against both the impersonal intrusion and key recovery attack simultaneously based on data-aware attack-detection.

Knowledge Based Pattern-Context-Aware Stereo Analysis and Its Applications

In this paper, we study a pattern-context-aware scheme for stereo pattern analysis.Depth and texture are chosen as two primary factors for the pattern-context- aware computing.We organize these patterns as a context to analyze.A knowledge-based inference system is built with human experience to model the correlation of the context and processing. The process for the pattern analysis could recognize potentially interested patterns by processing the optimal belief context. The strategy enables the system aware of the uncertainty in the pattern. An enhanced learning approach is introduced to allow the system to process the ambiguous pattern and to refine the confidence. An example is given to show the feasibility of the proposed scheme. It can be seen that the potential patterns can be differentiated and rebuilt as an extracted stereo model with the context-awareness. The discussion on potential applications for the intelligent driving systems is presented.

Fuzzy knowledge based enhanced matting

The goal of this paper is to address how to use human experience to develop an enhanced matting strategy. Based on a recursive α optimization framework, we present an adaptive fuzzy learning strategy for enhancement of matting. Taking into account the uncertainty of data, the proposed scheme successfully applies the expert human knowledge into matting. Experimental results are given to demonstrate the effect of the proposed method compared to some classical methods. The results indicate that the proposed adaptive learning algorithm handles uncertain pixels and perform stable matting.

Mixed Reality with a Collaborative Information System

In this paper, we present a mixed reality environment (MIXER) for immersive interactions. MIXER is an agent based collaborative information system displaying hybrid reality merging interactive computer graphics and real objects. MIXER is an agent based collaborative information system displaying hybrid reality merging interactive computer graphics and real objects. The system comprises a sensor subsystem, a network subsystem and an interaction subsystem. Related issues to the concept of mixed interaction, including human aware computing, mixed reality fusion, agent based systems, collaborative scalable learning in distributed systems, QoE-QoS balanced management and information security, are discussed. We propose a system architecture to perform networked mixed reality fusion for Ambient Interaction. The components of the mixed reality suit to perform human aware interaction are Interaction Space, Motion Monitoring, Action and Scenario Synthesisers, Script Generator, Knowledge Assistant Systems, Scenario Display, and a Mixed Reality Module. Thus, MIXER as an integrated system can provide a comprehensive human-centered mixed reality suite for advanced Virtual Reality and Augmented Reality applications such as therapy, training, and driving simulations.