Catálogo de publicaciones - libros

A kind of bisynchronous approach, including two processes of self-synchronization and additive synchronization, is proposed for audio watermarking to get the robust performance. The self-synchronous valley is created for watermark location rapidly, and the additive synchronous approach is adopted to locate the usable watermarks. Moreover, an adaptive matching approach is stated to obtain the final watermark from several candidates. The experimental results show that the valleys can be searched rapidly and effectively. The approach is robust against most common attacks.

This paper explains the work to design an information hiding based secure communication system, named covert speech telephone (CST). The overall system is designed over the internet using UDP protocol. Based on a GUI (graphical user interface) software, CST is possible to execute two optional secure modes. It is a completely digital system with high speech quality. Practical effects show that CST can meet the requirement of real-time secure communication. This new technique can effectively guarantee information security in VoIP system.

Temporal attacks will affect temporal synchronization signals loss. In this paper, we propose a low complexity temporal synchronization recovery method using local variance statistics in each group of picture (GOP), which is regarded as the feature parameters and sent as side information to recover synchronization signals. The temporal distortions can be identified by comparision of the feature parameters and the feature statistics of the received watermarked video data. Simulation results show that the proposed method is more robust against temporal attacks.

In this paper, an efficient video watermarking scheme is presented through modifying the third decoded luminance differential DC component in each selected macro block. The modification is implemented by binary dither modulation with adaptive quantization step. The proposed scheme is based on the observation that luminance differential DC components inside one macro block are generally space correlated, so the quantization step can be adjusted according to adjacent differential components, to utilize properties of human visual system (HVS). This method is very robust to gain attacks since amplitude scaling will have the same effect on differential components and the quantization step. Experimental results show that it can be implemented in real time with better visual quality than uniform-quantizing scheme.

We propose a novel block-based symmetric encryption system based on an of independently iterated chaotic logistic maps with global and local feedback as a diffusion process. Local feedback represents the temporal evolution of a single map, while global feedback represents the temporal evolution of the whole system (cross-map evolution). For security, the cryptosystem periodically modifies its internal configuration using a three-level random perturbation scheme, one at system-key (reset operation) and two at map array level (to increase the chaotic cycle length of the system). An analysis of the proposed scheme regarding its vulnerability to attacks, statistical properties and implementation performance is presented. To the best of our knowledge we provide a simple and secure scheme with the fastest software implementation reported in the literature.

We present an image authentication scheme based on feature extraction and codeword clustering in this paper. The two-level detections can be performed progressively based on the requirements of the authenticator. In order to generate these two-level authentication codes, the quad-tree segmentation and clustered VQ techniques are used. Using the first-level authentication code, the malicious tampered areas can be detected and located. In the second-level, it not only provides a more rigid detection of tampered areas but also can further remedy them. According to the experimental results, our scheme can correctly detect the malicious tampering and tolerate some incidental modifications, and then successfully remedy the tampered area. In addition, the space cost of authentication code is quite little.

We propose an intrinsic-distance based segmentation approach for segmenting human body parts in video frames. First, since the human body can be seen as a set of articulated parts, we utilize the moving articulated attributes to identify body part candidate regions automatically. The candidate regions and the background candidate regions are generated by voting and assigned to the spatiotemporal volume, which is comprised of frames of the video. Then, the intrinsic distance is used to estimate the boundaries of each body part. Our intrinsic distance-based segmentation technique is applied in the spatiotemporal volume to extract the optimal boundaries of the intrinsic distance in a video and obtain segmented frames from the segmented volume. The segmented results show that the proposed approach can tolerate incomplete and imprecise candidate regions because it provides temporal continuity. Furthermore, it can reduce over growing in the original intrinsic distance-based algorithm, since it can handle ambiguous pixels. We expect that this research can provide an alternative to segmenting a sequence of body parts in a video.

This paper proposes a kind of remote rendering method based on awareness model. This method takes the additional cost caused by the movement of the viewpoint into cost calculation and designs a cost prediction algorithm based on the vision field divided by awareness model. The simulation results show that the improved method can not only improve the quality of the remote rendering, but also make full use of the bandwidth of the network, as well as make the remote rendering more fluent when the viewpoint moves fast.

Sign language recognition (SLR) with large vocabulary and signer independency is valuable and is still a big challenge. Signer adaptation is an important solution to signer independent SLR. In this paper, we present a method of etyma-based signer adaptation for large vocabulary Chinese SLR. Popular adaptation techniques including Maximum Likelihood Linear Regression (MLLR) and Maximum A Posteriori (MAP) algorithms are used. Our approach can gain comparative results with that of using words, but we only require less than half data.

We propose a novel news browsing system that can cluster photo news articles based on both textual features of articles and image features of news photos for a personal news database which is built by accumulating Web photo news articles. The system provides two types of clustering methods: normal clustering and thread-style clustering. It enables us to browse news articles over several weeks or months visually and find out useful news easily. In this paper, we describe an overview of our system, some examples of uses and user studies.