Kernel-based learning algorithms have been successfully applied in various problem domains, given appropriate kernel functions. In this paper, we discuss the problem of designing kernel functions for binary regression and show that using a bell-shaped cosine function as a kernel function is optimal in some sense. The rationale of this result is based on the Karhunen-Loeve expansion, i.e., the optimal approximation to a set of functions is given by the principal component of the correlation operator of the functions.

The EDD connection admission control scheme has been proposed for supporting real-time communication in packet-switched networks. In the scheme, when a connection establishment request occurs, the worst-case link delay in each link along the connection is calculated to determine whether the request can be accepted or not. In order to calculate the worst-case link delay, we must perform a check called the point schedulability check for each of some discrete time instants (checkpoints). Therefore when there are many checkpoints, the worst-case link delay calculation is time-consuming. We have proposed a high-speed calculation method. The method finds some checkpoints for which the point schedulability check need not be performed and removes such unnecessary checkpoints in advance before a connection establishment request occurs, and the check is performed for each of the remaining checkpoints after the request occurs. However, the method is not so effective under the situation that the maximum packet length in networks is large, because the method can find few unnecessary checkpoints under the situation. This paper proposes a new high-speed calculation method. We relax the condition which determines whether or not the point schedulability check need not be performed for each checkpoint in our previous method and derive a new condition for finding unnecessary checkpoints. Using the proposed method based on the new condition, we can increase the number of unnecessary checkpoints compared to our previous method. Numerical examples which are obtained by extensive simulation show that the proposed method can attain as much as about 50 times speedup.

In the present paper, we propose a method for reconstructing the surfaces of objects from stereo data. Both the fitness of stereo data to surfaces and interrelation between the surfaces are defined in the framework of a three-dimensional (3-D) Markov Random Field (MRF) model. The surface reconstruction is accomplished by searching for the most likely state of the MRF model. Three experimental results are shown for synthetic and real stereo data.

This paper proposes a gonio-spectral imaging system for measuring light reflection on an object surface by using two robot arms, a multi-band lighting system, and a monochrome digital camera. It allows four degrees of freedom in incident and viewing angles necessary for full parametrization of a reflection model function. Spectral images captured for various incident and viewing angles are warped as if they were all captured from the same viewing direction. The intensity of reflected light is thus recorded in a normalized image form for any incident and viewing directions. The normalized images are used to estimate reflection model parameters at each surface point. To ensure point-wise reflection modeling, a calibration method is also proposed based on a geometrical model of the robot arms and camera. The proposed system can deal with objects with surface texture. Experiments are done on system calibration, reflection model, and spectral estimation. The results using colored objects show the feasibility of the proposed imaging system.

Natural, continuous tone images have a very important property of high correlation of adjacent pixels. Images which we wish to compress are usually non-stationary and can be reasonably modeled as smooth and textured areas separated by edges. This property has been successfully exploited in LOCO-I and CALIC by applying gradient based predictive coding as a major de-correlation tool. However, they only examine the horizontal and vertical gradients, and assume the local edge can only occur in these two directions. Their over-simplified assumptions hurt the robustness of the prediction in higher complex areas. In this paper, we propose an accurate gradient selective prediction (AGSP) algorithm which is designed to perform robustly around any type of image texture. Our method measures local texture information by comparison and selection of normalized scalar representation of the gradients in four directions. An adaptive predictor is formed based on the local gradient information and immediate causal pixels. Local texture properties are also exploited in the context modeling of the prediction error. The results we obtained on a test set of several standard images are encouraging. On the average, our method achieves a compression ratio significantly better than CALIC without noticeably increasing of computational complexity.

In this paper, we present a method for recognition of human activity as a series of actions from an image sequence. The difficulty with the problem is that there is a chicken-egg dilemma that each action needs to be extracted in advance for its recognition but the precise extraction is only possible after the action is correctly identified. In order to solve this dilemma, we use as many models as actions of our interest, and test each model against a given sequence to find a matched model for each action occurring in the sequence. For each action, a model is designed so as to represent any activity containing the action. The hierarchical hidden Markov model (HHMM) is employed to represent the models, in which each model is composed of a submodel of the target action and submodels which can represent any action, and they are connected appropriately. Several experimental results are shown.

This paper presents a 3D feature-based binocular tracking algorithm for tracking crowded people indoors. The algorithm consists of a two stage 3D feature points grouping method and a robust 3D feature-based tracking method. The two stage 3D feature points grouping method can use kernel-based ISODATA method to detect people accurately even though the part or almost full occlusion occurs among people in surveillance area. The robust 3D feature-based Tracking method combines interacting multiple model (IMM) method with a cascade multiple feature data association method. The robust 3D feature-based tracking method not only manages the generation and disappearance of a trajectory, but also can deal with the interaction of people and track people maneuvering. Experimental results demonstrate the robustness and efficiency of the proposed framework. It is real-time and not sensitive to the variable frame to frame interval time. It also can deal with the occlusion of people and do well in those cases that people rotate and wriggle.

An effective human-robot interaction is essential for wide penetration of service robots into the market. Such robot needs a vision system to recognize objects. It is, however, difficult to realize vision systems that can work in various conditions. More robust techniques of object recognition and image segmentation are essential. Thus, we have proposed to use the human user's assistance for objects recognition through speech. This paper presents a system that recognizes objects in occlusion and/or multicolor cases using geometric and photometric analysis of images. Based on the analysis results, the system makes a hypothesis of the scene. Then, it asks the user for confirmation by describing the hypothesis. If the hypothesis is not correct, the system generates another hypothesis until it correctly understands the scene. Through experiments on a real mobile robot, we have confirmed the usefulness of the system.

A new class of ternary sequence with a zero-correlation zone is introduced. The proposed sequence sets have a zero-correlation zone for both periodic and aperiodic correlation functions. The proposed sequences can be constructed from a pair of Hadamard matrices of size n0n0 and a Hadamard matrix of size n1n1. The constructed sequence set consists of n0 n1 ternary sequences, and the length of each sequence is (n1+1) for a non-negative integer m. The zero-correlation zone of the proposed sequences is |τ|≤ -1, where τ is the phase shift. The sequence member size of the proposed sequence set is equal to times that of the theoretical upper bound of the member size of a sequence set with a zero-correlation zone.

Time-frequency-selective, i.e., time-variant multipath, fading in orthogonal frequency division multiplexing (OFDM) systems destroys subcarrier orthogonality, resulting in intercarrier interference (ICI). In general, the previously proposed estimation schemes to resolve this problem are only applicable to slowly time-variant channels or suffer from high complexity due to large-sized matrix inversion. In this letter, we propose and develop efficient symbol estimation schemes, called the iterative sequential neighbor search (ISNS) algorithm and the simplified iterative sequential neighbor search (S-ISNS) algorithm. These algorithms achieve enhanced performances with low complexities, compared to the existing estimation methods.

In this paper, a chromatic adaptation model (CAM) based on the dependence on LMS cone responses of the human visual system (HVS) is proposed for TV and PC monitors under a variety of viewing conditions. We derived the proposed CAM based on Breneman's corresponding color data. The results of the experiments were carried out to assess the proposed model performance in terms of color fidelity by comparing complex images on a LCD monitor. We confirmed that the proposed model performed better to predict corresponding colors under various viewing conditions. Therefore, the reproduced colors, which are viewed in real surround viewing conditions, are perceived the same as original object colors, when the proposed CAM was applied to color display devices such as CRT, LCD, and PDP.

This paper proposes a simple and efficient method to numerically obtain the mapping degree deg(f, 0, B) of a C1 map f : Rn → Rn at a regular value 0 relative to a bounded open subset B ⊂ Rn. For practical application, this method adopts Aberth's algorithm which does not require computation of derivatives and determinants, and reduces the computational cost with two additional procedures, namely preconditioning using the coordinate transformation and pruning using Krawczyk's method. Numerical examples show that the proposed method gives the mapping degree with 2n+1 operations using interval arithmetic.

This paper presents a new statistical model-based voice activity detection (VAD) algorithm in the wavelet domain to improve the performance in non-stationary environments. Due to the efficient time-frequency localization and the multi-resolution characteristics of the wavelet representations, the wavelet transforms are quite suitable for processing non-stationary signals such as speech. To utilize the fact that the wavelet packet is very efficient approximation of discrete Fourier transform and has built-in de-noising capability, we first apply wavelet packet decomposition to effectively localize the energy in frequency space, use spectral subtraction, and employ matched filtering to enhance the SNR. Since the conventional wavelet-based spectral subtraction eliminates the low-power speech signal in onset and offset regions and generates musical noise, we derive an improved multi-band spectral subtraction. On the other hand, noticing that fixed threshold cannot follow fluctuations of time varying noise power and the inability to adapt to a time-varying environment severely limits the VAD performance, we propose a statistical model-based VAD algorithm in wavelet domain with an adaptive threshold. We perform extensive computer simulations and compare with the conventional algorithms to demonstrate performance improvement of the proposed algorithm under various noise environments.

Recent work has shown that the usage of multiple antennas at the transmitter and receiver in a flat fading environment results in a linear increase in channel capacity. But increasing the number of antennas induces the higher hardware costs and computational burden. To overcome those problems, it is effective to select antennas appropriately among all available ones. In this paper, a new antenna selection method is proposed. The transmit antennas are selected so as to maximize the channel capacity using the genetic algorithm (GA) which is the one of the general random search algorithm. The results show that the proposed GA achieves almost the same performance as the optimal selection method with less computational amount.

This paper describes a -90 dBc@10 kHz phase noise fractional-N frequency synthesizer of 110 M-180 MHz output with accurate loop bandwidth control. Stable phase noise characteristics are achieved by controlling the bandwidth correctly, even if the PLL uses a noisy but small ring oscillator. Digital controller adjusts voltage controlled oscillator (VCO) gain and time constant of the loop filter. Analog controller compensates temperature variance. Test chip fabricated on 0.13 µm CMOS process shows stable and 6.8 dB improvement of the phase noise performance is achieved against process and environmental variations.

A spread-spectrum clock generator (SSCG) using fractional-N phase-locked loop (PLL) with an extended range sigma-delta (ΣΔ) modulator is presented in this paper. The proposed ΣΔ modulator simply adds an extra output bit in the first stage modulator. It can enlarge the input range about three times as compared to the conventional modulator and solve the saturation problem when the input exceeds the boundary of the conventional modulator. A flexible digital modulation controller can generate center and down spread-spectrum modulation and each has spread ratios of 0.4%, 0.8%, 1.6% and 3.2%. The proposed SSCG has been fabricated in TSMC 0.35-µm double-poly quadruple-metal CMOS process with output frequency of 300 MHz. The active area is 0.630.62 mm2 and the power consumption is 17.5 mW.

In this paper, we present a polling scheme which allows for augmenting the support of voice communications in point co-ordination function (PCF) of IEEE 802.11 wireless networks. In this scheme, the Access Point (AP) of the Basic Service Set (BSS) maintains two polling lists, i.e. the talking list and the silence list. Based on the talking status of the stations identified via silence detection, two lists are dynamically adjusted by the AP. Temporary removal is applied to the stations in the silence list to further upgrade the performance. The conducted study based on simulation has shown that the proposed scheme can support more voice stations and has a lower packet loss rate than that obtained by four reference polling algorithms.

In this paper, we propose a combination-based novel technique of dithered subband coding with spectral subtraction for improving the perceptual quality of coded audio at low bit rates. It is well known that signal-correlated distortion is audible when the audio signal is quantized at bit rates lower than the lower bound of perceptual coding. We show that this problem can be overcome by applying the dithering quantization process in each subband. Consequently, the quantization noise is rendered into a signal-independent white noise; this noise is then estimated and removed by spectral subtraction at the decoder. Experimental results show an effective improvement by the proposed method over the conventional one in terms of better SNR and human listening test results. The proposed method can be combined with other existing or future coding methods such as perceptual coding to improve their performance at low bit rates.

Usually an FIR filter is used to model the physical paths in an active noise control system. However, the order of the filter to be modeled is a key factor for determining the computational load for the adaptive algorithms associated with active noise control (ANC), particularly for multi-channel algorithms. In this letter, the relationships among the filter's order, the plant modeling error and the location of poles for the transfer functions of the physical paths in an ANC system are theoretically examined and numerical examples are given to verify the theoretical results.

We investigated physical conditions for optimum display systems on various TV viewing conditions, and found that visual brightness function could be derived from relationships between Steven's power law and Bartleson-Breneman's brightness function, and that the optimum physical contrast ratio and compensated gamma for display system with adaptation luminance level could be obtained from the proposed brightness function.