Fumihiko SAKAUE Takeshi SHAKUNAGA
The present paper reports a robust projection onto eigenspace that is based on iterative projection. The fundamental method proposed in Shakunaga and Sakaue and involves iterative analysis of relative residual and projection. The present paper refines the projection method by solving linear equations while taking noise ratio into account. The refinement improves both the efficiency and robustness of the projection. Experimental results indicate that the proposed method works well for various kinds of noise, including shadows, reflections and occlusions. The proposed method can be applied to a wide variety of computer vision problems, which include object/face recognition and image-based rendering.
Tae Hoon LEE Won Sang RA Seung Hee JIN Tae Sung YOON Jin Bae PARK
A new robust extended Kalman filter is proposed for the discrete-time nonlinear systems with norm-bounded parameter uncertainties. After linearization of the nonlinear systems, the uncertainties described by the energy bounded constraint can be converted into an indefinite quadratic cost function to be minimized. The solution to the minimization problem is given by the extended Kalman filter derived in a Krein space, which leads to a robust version of the extended Kalman filter. Since the resulting robust filter has the same structure as a standard extended Kalman filter, the proposed filter can be readily designed by simply including the uncertainty terms in its formulas. The results of simulations are presented to demonstrate that the proposed filter achieves the robustness against parameter variation and performs better than the standard extended Kalman filter.
Chih-Peng HUANG Shi-Ting WANG Yau-Tarng JUANG
This paper presents a distinct approach to the robustness stability analysis and design of linear uncertain systems. Based on the extension version of the projection method, the specific stability issue, which ensures the poles within a specific region, can be efficiently analyzed. Furthermore, we derive a simple design scheme for a class of uncertain systems. By the proposed numerical algorithm, some examples are given to demonstrate the validity and effectiveness.
Yoshio KONNO Jianting CAO Takayuki ARAI Tsunehiro TAKEDA
Treating an averaged multiple-trials data or non-averaged single-trial data is a main approach in recent topics on applying independent component analysis (ICA) to neurobiological signal processing. By taking an average, the signal-to-noise ratio (SNR) is increased but some important information such as the strength of an evoked response and its dynamics will be lost. The single-trial data analysis, on the other hand, can avoid this problem but the SNR is very poor. In this study, we apply ICA to both non-averaged single-trial data and averaged multiple-trials data to determine the properties and advantages of both. Our results show that the analysis of averaged data is effective for seeking the response and dipole location of evoked fields. The non-averaged single-trial data analysis efficiently identifies the strength and dynamic component such as α-wave. For determining both the range of evoked strength and dipole location, an analysis of averaged limited-trials data is better option.
In this paper our recently introduced method called output distributional influence function (ODIF) is used for the evaluation of the robustness properties of the nonlinear filter class of morphological filters. Several examples of the ODIFs for the dilation, closing and clos-opening are given and explained carefully. For each of these morphological filters the effect of filter length is examined by using the ODIFs for the expectation and variance. The robustness properties of the filters are also compared to each other and the effect of the distribution of the contamination is investigated for the closing as an example of realistic filtering conditions.
Jun YANG Yew-Hin LIEW Woon-Seng GAN
This letter outlines a scheme to produce a wider robust bandwidth, with better approximations to the perfect reproduction of pre-recorded acoustic signals. Multi-parameter inverse filtering method is proposed in the virtual sound imaging system for improving the robustness performance. The superiority of this new type of inverse filter is demonstrated on a 3-speaker system.
Yasuyuki SUGAYA Kenichi KANATANI
Many feature tracking algorithms have been proposed for motion segmentation, but the resulting trajectories are not necessarily correct. In this paper, we propose a technique for removing outliers based on the knowledge that correct trajectories are constrained to be in a subspace of their domain. We first fit an appropriate subspace to the detected trajectories using RANSAC and then remove outliers by considering the error behavior of actual video tracking. Using real video sequences, we demonstrate that our method can be applied if multiple motions exist in the scene. We also confirm that the separation accuracy is indeed improved by our method.
Miheung CHOE Hyunduk KANG Kiseon KIM
To sample a band-limited analog signal directly from the high frequency down to the baseband for the digital signal processing with significantly reduced computation, several concepts of the bandpass sampling are introduced. In this paper, a robust bandpass sampling scheme when there exist frequency deviations due to the channel effect and hardware instability is proposed for practical use, and the effects of the frequency deviations are discussed to select a proper sampling frequency.
The alternative c-means algorithm has recently been presented by Wu and Yang for robust clustering of data. In this letter, we analyze the convergence of this algorithm by transforming it into an equivalent form with the Legendre transform. It is shown that this algorithm converges to a local optimal solution from any starting point.
Konstantin MARKOV Tomoko MATSUI Rainer GRUHN Jinsong ZHANG Satoshi NAKAMURA
This paper presents the ATR speech recognition system designed for the DARPA SPINE2 evaluation task. The system is capable of dealing with speech from highly variable, real-world noisy conditions and communication channels. A number of robust techniques are implemented, such as differential spectrum mel-scale cepstrum features, on-line MLLR adaptation, and word-level hypothesis combination, which led to a significant reduction in the word error rate.
For an uncertain discrete event system (DES) modeled as a set of some possible nondeterministic automata, we address robust supervisory control problems. Based on language models, this paper presents the existence conditions of a robust nonblocking (RN) supervisor that guarantees the absence of blocked states in a closed-loop system. We show that an RN supervisor achieves both a given language specification and the nonblocking characteristics of any nondeterministic automata of the set.
In this letter, we propose a two-stage object-based error-concealment technique for MPEG-2 video transmitted in a burst-packet-loss environment. A burst packet loss typically destroys a large area of MPEG-2 coded video. In the first stage, the missing area is intra-picture estimated and concealed in terms of a region-based approach. In the second stage, irregular-moving objects with respect to the background are identified and compensated from the predictive picture. As compared with conventional concealment approaches, the proposed method achieves better PSNR performance and reduces the visual artifacts.
Montree BUDSABATHON Shinsuke HARA
In this paper, we present the theoretical analysis of the bit error rate (BER) performance of Single-Carrier Modulation (SCM) and Orthogonal Frequency Division Multiplexing (OFDM) systems under two types of temporally localized man-made noises (generalized shot noise and bursty noise models) environments. The robustness of OFDM system against these two kinds of man-made noises is discussed and then compared with that of SCM system at the same transmission rate. We show that for OFDM system, the BER performance highly depends on the number of subcarriers and the strength of the man-made noise, i.e., the level of the power spectral density of the man-made noise. In addition to the common knowledge on OFDM, we show that OFDM system is sometime less robust to the man-made noises than SCM system.
Lae-Hoon KIM Jun-Seok LIM Koeng-Mo SUNG
In loudspeaker-based 3D audio systems, there are some acoustic crosstalk cancellation methods to enlarge the 'sweet spot' around a fixed listener position. However, these methods have common defect that most of them can be applied only to the specific narrow frequency band. In this letter, we propose the more robust acoustic crosstalk cancellation method so that we can cancel the crosstalk signal in far wider frequency band and enlarge 'sweet spot. ' For this goal, we apply a sum and difference filter to the conventional three loudspeaker-based 3D audio system.
Kazuyuki TAKAGI Rei OGURO Kazuhiko OZEKI
Experiments were conducted to examine an approach from language modeling side to improving noisy speech recognition performance. By adopting appropriate word strings as new units of processing, speech recognition performance was improved by acoustic effects as well as by test-set perplexity reduction. Three kinds of word string language models were evaluated, whose additional lexical entries were selected based on combinations of part of speech information, word length, occurrence frequency, and log likelihood ratio of the hypotheses about the bigram frequency. All of the three word string models reduced errors in broadcast news speech recognition, and also lowered test-set perplexity. The word string model based on log likelihood ratio exhibited the best improvement for noisy speech recognition, by which deletion errors were reduced by 26%, substitution errors by 9.3%, and insertion errors by 13%, in the experiments using the speaker-dependent, noise-adapted triphone. Effectiveness of word string models on error reduction was more prominent for noisy speech than for studio-clean speech.
The technique of the digital watermarking is one of the ways to resolve copyright ownership and verify originality of digital contents (e.g. text documents, audio, still images, video, etc.). In this paper, we obtained global robustness and minimal error through using frame based watermarking and including DC and AC coefficients of DCT transform, which extended conventional watermarking method having local robustness and error through using block based and/or AC coefficients only. As a result, the high robustness and quality of our method were proved by several attacks such as lossy image compression, linear filtering, additive noise, scaling, cropping and so on. Watermarks embedded by our method are survived most of JPEG compressions.
Daniel FOGARAS Kokichi SUGIHARA
The paper presents a topology-oriented robust algorithm for the incremental construction of line arrangements. In order to achieve a robust implementation, the topological and geometrical computations are strictly separated. The topological part is proved to be reliable without any assumption on the accuracy of the geometrical part. A self-correcting property is introduced to minimize the effect of numerical errors. Computational experiments show how the self-correcting property works, and we also discuss some applications of the algorithm.
Shih-Chang HSIA I-Chang JOU Shing-Ming HWANG
Watermarking techniques are widely used to protect the secret document. In some valuable literatures, most of them concentrate on the binary data watermarking by using comparisons of an original image and a watermarked image to extract the watermark. In this paper, an efficient watermarking algorithm is presented with two-layer hidden for gray-level image watermarking. In the first layer, the key information is found based on the codebook concept. Then the secret key is further hidden to the watermarked image adopting the encryption consisting of spatial distribution in the second layer. The simulations demonstrate that the watermarking information is perceptually invisible in the watermarked image. Moreover, the gray-level watermark can be extracted by referring key parameters rather than the original image, and the extracting quality is very good.
This paper focuses on the watermarking system using a controlled quantization process. We first present a model of the watermark embedding and extracting processes and carry out their analyses. Then we examine the robustness of the watermarking system against common image processing and clarify the reason why detection errors occur in the watermark extracting process. Based on the result, we improve the watermark extracting process and design robust watermarking systems. The improvement is accomplished using a deconvolution filter and neural network techniques. Numerical experiments using the DCT-based watermarking system show good performance as expected by us.
Byung-Gun PARK Wook HYUN KWON Jae-Won LEE
This paper proposes a receding horizon control scheme for a set of uncertain discrete-time linear systems with randomly jumping parameters described by a finite-state Markov process whose jumping transition probabilities are assumed to belong to some convex sets. The control scheme for the underlying systems is based on the minimization of an upper bound on the worst-case infinite horizon cost function at each time instant. It is shown that the mean square stability of the proposed control system is guaranteed under some matrix inequality conditions on the terminal weighting matrices. The proposed controller is obtained using semidefinite programming.