In this letter, we elucidate the ergodic capacity of multiple-input multiple-output (MIMO) systems with M-ary phase-shift keying (MPSK) modulation and time-multiplexed pilots in frequency-flat Rayleigh fading environment. With linear minimum mean square error (LMMSE) channel estimation, the optimal pilots design is presented. For mathematical tractability, we derive an easy-computing closed-form lower bound of the channel capacity. Based on the lower bound, the optimal power allocation between the data and pilots is also presented in closed-form, and the optimal training length is investigated by numerical optimization. It is shown that the transmit scheme with equal training and data power and optimized training length provides suboptimal performance, and the transmit scheme with optimized training length and training power is optimal. With the latter scheme, in most situations, the optimal training length equals the number of the transmit antennas and the corresponding optimal power allocation can be easily computed with the proposed formula.

This paper presents an alternative traffic model for an ATM multiplexer providing video, voice, image, and data services. The traffic model classifies the input traffic into two types: real-time and non-real-time. The input process for realtime traffic is periodic and correlated, while that for non-realtime traffic is batch Poisson and independent. This multiplexer is assumed to be a priority queueing system with synchronous servers operating on time-frame basis and with separate finite buffers for each type of traffic. State probabilities and performance measures are successfully obtained using a Markov analysis technique and an application of the residue theorem in complex variable. The results can be applied in the design of an ATM multiplexer.

This letter addresses the problem of robust transceiver design for the multiuser multiple-input-multiple-output (MIMO) downlink where the channel state information at the base station (BS) is imperfect. A stochastic approach which minimizes the expectation of the total mean square error (MSE) of the downlink conditioned on the channel estimates under a total transmit power constraint is adopted. The iterative algorithm reported in [2] is improved to handle the proposed robust optimization problem. Simulation results show that our proposed robust scheme effectively reduces the performance loss due to channel uncertainties and outperforms existing methods, especially when the channel errors of the users are different.

In this letter, we discuss the problem of receive antenna selection in the downlink of multiuser multiple-input multiple-output (MIMO) systems with Tomlinson-Harashima precoding (THP), where the number of receivers is assumed equal to that of transmit antennas. Based on the criterion of maximum system sum-capacity, a per-layer receive antenna selection scheme is proposed. This scheme, which selects one receive antenna for each receiver, can well exploit the nonlinear and successive characteristics of THP. Two models are established for the proposed per-layer scheme and the conventional per-user scheme. Both the theoretical analysis and simulation results indicate that the proposed scheme can greatly improve the equivalent channel power gains and the system sum-capacity.

Spatial multiplexing (SM) offers a linear increase in transmission rate without bandwidth expansion or power increase. In SM systems, the LMMSE receiver establishes a good tradeoff between the complexity and performance. The performance of the LMMSE receiver would be degraded by MIMO channel estimation errors. This letter focus on obtaining the asymptotic convergence of output interference power and SIR performance for the LMMSE receiver with channel uncertainty. Exactly matched simulation results verify the validity of analysis in the large-system assumption. Furthermore, we find that the analytical results are also valid in the sense of average results for limited-scale system in spite of the asymptotic assumption used in derivation.

This paper investigates a new method for fusion of scores generated by multiple classification sub-systems that help to further reduce the classification error rate in Spoken Language Recognition (SLR). In recent studies, a variety of effective classification algorithms have been developed for SLR. Hence, it has been a common practice in the National Institute of Standards and Technology (NIST) Language Recognition Evaluations (LREs) to fuse the results from several classification sub-systems to boost the performance of the SLR systems. In this work, we introduce a discriminative performance measure to optimize the performance of the fusion of 7 language classifiers developed as IIR's submission to the 2009 NIST LRE. We present an Error Corrective Fusion (ECF) method in which we iteratively learn the fusion weights to minimize error rate of the fusion system. Experiments conducted on the 2009 NIST LRE corpus demonstrate a significant improvement compared to individual sub-systems. Comparison study is also conducted to show the effectiveness of the ECF method.

The paper studies a high-ranking node in a broadband integrated services digital network(B-ISDN). The input traffic is classified into two types: real-time and non-real-time. For each type of input traffic, we assume that the message arrival process is a batch Poisson process and that the message size is arbitrarily distributed so as to describe services from narrowband to wideband. We model the high-ranking node by a queueing system with multiple synchronous servers and two separate finite buffers, one for each type of traffic. We derive performance measures exactly by using a two-dimensional imbedded discrete-time Markov chain analysis, within which the transition probabilities are obtained via an application of the residue theorem in complex variables. The performance measures include the blocking probability, delay, and throughput.

Downlink multiuser MIMO system has attracted considerable attention recently for its potential to increase the system capacity. However, due to the limitation on the number of transmit antennas, when there are more users than can be supported simultaneously in a cell, other multiple access schemes, such as TDMA, must be applied in combination with multiuser MIMO. In this paper, we aim to design practical user scheduling algorithms to maximize the system capacity. Because the brute-force search for optimal user allocation is computationally prohibitive, we propose three low complexity suboptimal scheduling algorithms that offer both low complexity and high performance.

Cognitive radio technology, which allows secondary user (SU) to utilize the spectrum holes left by primary user (PU), was proposed to solve spectrum under-utilization problem. However, due to sensing error, SU's transmission will bring negative effects to PU's communication. Recently, cooperative relay technology was introduced to solve this problem. In this paper, a cooperative framework, which allows SU to act as a relay for primary link when needed, is considered and then a cognitive relay scheme is proposed. In order to maximize SU's throughput while keeping the system stable, we study and obtain SU's optimal strategy (i.e., relaying strategy and power allocation) by a constrained optimization problem. Since energy consumption is also an important problem for cognitive radio networks, we also investigate SU's optimal strategy to maximize SU's energy efficiency while keeping the system stable. The numerical results show that the cognitive relay scheme can achieve higher throughput and energy efficiency than reference schemes.

In this letter, coordinated power allocation (PA) is investigated for the downlink of a generalized multi-cluster distributed antenna system (DAS). Motivated by practical applications, we assume only the global large-scale channel state information is known at the transmitter. First, an upper bound (UB) for the ergodic sum capacity of the system is derived and used as a simplified optimization target. Then, a coordinated PA scheme is proposed based on Geometric Programming (GP), which is demonstrated to be nearly optimal by Monte Carlo simulations.

Ensemble learning is widely used in the field of sensor network monitoring and target identification. To improve the generalization ability and classification precision of ensemble learning, we first propose an approximate attribute reduction algorithm based on rough sets in this paper. The reduction algorithm uses mutual information to measure attribute importance and introduces a correction coefficient and an approximation parameter. Based on a random sampling strategy, we use the approximate attribute reduction algorithm to implement the multi-modal sample space perturbation. To further reduce the ensemble size and realize a dynamic subset of base classifiers that best matches the test sample, we define a similarity parameter between the test samples and training sample sets that takes the similarity and number of the training samples into consideration. We then propose a k-means clustering-based dynamic ensemble selection algorithm. Simulations show that the multi-modal perturbation method effectively selects important attributes and reduces the influence of noise on the classification results. The classification precision and runtime of experiments demonstrate the effectiveness of the proposed dynamic ensemble selection algorithm.

This paper considers tracking of a non-cooperative emitter based on a single sensor. To this end, the direct target motion analysis (DTMA) approach, where the target state is straightforwardly achieved from the received signal, is exploited. In order to achieve observability, the sensor has to perform a maneuver relative to the emitter. By suitably building an approximated likelihood function, the unscented Kalman filter (UKF), which is able to work under high nonlinearity of the measurement model, is adopted to recursively estimate the target state. Besides, the posterior Cramér-Rao bound (PCRB) of DTMA, which can be used as performance benchmark, is also achieved. The effectiveness of proposed method is verified via simulation experiments.

Vector precoding is a nonlinear broadcast precoding scheme in the downlink of multi-user MIMO systems which outperforms linear precoding and THP (Tomlinson-Harashima Precoding). This letter discusses the problem of joint receive antenna selection in the multi-user MIMO downlink with vector precoding. Based on random matrix analysis, we derive a simple heuristic selection criterion using singular value decomposition (SVD) and carry out an exhaustive search to determine for each user which receive antenna should be used. Simulation results reveal that receive antenna selection using our proposed criterion obtains the same diversity order as the optimal selection criterion.

In this paper, we propose integration of multimodal features using conditional random fields (CRFs) for the segmentation of broadcast news stories. We study story boundary cues from lexical, audio and video modalities, where lexical features consist of lexical similarity, chain strength and overall cohesiveness; acoustic features involve pause duration, pitch, speaker change and audio event type; and visual features contain shot boundaries, anchor faces and news title captions. These features are extracted in a sequence of boundary candidate positions in the broadcast news. A linear-chain CRF is used to detect each candidate as boundary/non-boundary tags based on the multimodal features. Important interlabel relations and contextual feature information are effectively captured by the sequential learning framework of CRFs. Story segmentation experiments show that the CRF approach outperforms other popular classifiers, including decision trees (DTs), Bayesian networks (BNs), naive Bayesian classifiers (NBs), multilayer perception (MLP), support vector machines (SVMs) and maximum entropy (ME) classifiers.

Recently, fuzzy set theory has been widely employed in building portfolio selection models where uncertainty plays a role. In these models, future security returns are generally taken for fuzzy variables and mathematical models are then built to maximize the investment profit according to a given risk level or to minimize a risk level based on a fixed profit level. Based on existing works, this paper proposes a portfolio selection model based on fuzzy birandom variables. Two original contributions are provided by the study: First, the concept of technical analysis is combined with fuzzy set theory to use the security returns as fuzzy birandom variables. Second, the fuzzy birandom Value-at-Risk (VaR) is used to build our model, which is called the fuzzy birandom VaR-based portfolio selection model (FBVaR-PSM). The VaR can directly reflect the largest loss of a selected case at a given confidence level and it is more sensitive than other models and more acceptable for general investors than conventional risk measurements. To solve the FBVaR-PSM, in some special cases when the security returns are taken for trapezoidal, triangular or Gaussian fuzzy birandom variables, several crisp equivalent models of the FBVaR-PSM are derived, which can be handled by any linear programming solver. In general, the fuzzy birandom simulation-based particle swarm optimization algorithm (FBS-PSO) is designed to find the approximate optimal solution. To illustrate the proposed model and the behavior of the FBS-PSO, two numerical examples are introduced based on investors' different risk attitudes. Finally, we analyze the experimental results and provide a discussion of some existing approaches.

This paper presents a novel acoustic modeling technique of large vocabulary automatic speech recognition for under-resourced languages by leveraging well-trained acoustic models of other languages (called source languages). The idea is to use source language acoustic model to score the acoustic features of the target language, and then map these scores to the posteriors of the target phones using a classifier. The target phone posteriors are then used for decoding in the usual way of hybrid acoustic modeling. The motivation of such a strategy is that human languages usually share similar phone sets and hence it may be easier to predict the target phone posteriors from the scores generated by source language acoustic models than to train from scratch an under-resourced language acoustic model. The proposed method is evaluated using on the Aurora-4 task with less than 1 hour of training data. Two types of source language acoustic models are considered, i.e. hybrid HMM/MLP and conventional HMM/GMM models. In addition, we also use triphone tied states in the mapping. Our experimental results show that by leveraging well trained Malay and Hungarian acoustic models, we achieved 9.0% word error rate (WER) given 55 minutes of English training data. This is close to the WER of 7.9% obtained by using the full 15 hours of training data and much better than the WER of 14.4% obtained by conventional acoustic modeling techniques with the same 55 minutes of training data.

Conventional features for Automatic Speech Recognition and Sound Event Recognition such as Mel-Frequency Cepstral Coefficients (MFCCs) have been shown to perform poorly in noisy conditions. We introduce an auditory feature based on the gammatone filterbank, the Selective Gammatone Envelope Feature (SGEF), for Robust Sound Event Recognition where channel selection and the filterbank envelope is used to reduce the effect of noise for specific noise environments. In the experiments with Hidden Markov Model (HMM) recognizers, we shall show that our feature outperforms MFCCs significantly in four different noisy environments at various signal-to-noise ratios.

Based on the minimum mean square error (MMSE) detection with iterative soft interference cancellation (SoIC), we propose an adaptive MMSE (A-MMSE) algorithm which acts as an MMSE operator at the beginning of iteration and a maximum ratio combination (MRC) when the interference is nearly cancelled. In our algorithm, a modified metric matrix based on the reliability of soft information from the decoder output is multiplied by the interference part of channel correlation matrix to update the detection operator. The simulation results have shown that this A-MMSE iterative SoIC algorithm can achieve significant performance advantage over the traditional MMSE iterative SoIC algorithm.

In an Multi-Protocol Label Switching (MPLS) network domain, Asynchronous Transfer Mode--Label Switch Routers (ATM-LSRs) are considered the best candidate for providing the highest forwarding capability. ATM-LSRs implement a VC-merging scheme that allows many IP routes to be mapped into the same VPI/VCI label, hence supporting scalability. The VC-merging requires reassembly buffers to reconstruct an original packet from its segmented but interleaved AAL-5 cells. In this paper, we analyze the performance of an ATM-LSR with partial VC-merging capability and investigate the impact of VC-merging on the requirement of the reassembly and output buffer. The numerical computation complexity of the mathematical analysis can be reduced from O(M4) to O(M2), where M is the total number of ON-OFF sources. We also propose a closed-form equation, which approximates the distribution of the output buffer with satisfactory accuracy. Numerical results show that when incoming cells are severely interleaved, the VC-merging needs the reassembly buffer size to be of the same order as the output buffer size, which cannot be ignored.