In this paper, we propose a joint transmit and receive linear filtering based on minimum mean square error criterion (joint Tx/Rx MMSE filtering) for single-carrier (SC) multiple-input multiple-output (MIMO) transmission. Joint Tx/Rx MMSE filtering transforms the MIMO channel to the orthogonal eigenmodes to avoid the inter-antenna interference (IAI) and performs MMSE based transmit power allocation to sufficiently suppress the inter-symbol interference (ISI) resulting from the severe frequency-selectivity of the channel. Rank adaptation and adaptive modulation are jointly introduced to narrow the gap of received signal-to-interference plus noise power ratio (SINR) among eigenmodes. The superiority of the SC-MIMO transmission with joint Tx/Rx MMSE filtering and joint rank adaptation/adaptive modulation is confirmed by computer simulation.

Recently, video surveillance systems have been widely introduced in various places, and protecting the privacy of objects in the scene has been as important as ensuring security. Masking each moving object with a background subtraction method is an effective technique to protect its privacy. However, the background subtraction method is heavily affected by sunshine change, and a redundant masking by over-extraction is inevitable. Such superfluous masking disturbs the quality of video surveillance. In this paper, we propose a moving object masking method combining background subtraction and machine learning based on Real AdaBoost. This method can reduce the superfluous masking while maintaining the reliability of privacy protection. In the experiments, we demonstrate that the proposed method achieves about 78-94% accuracy for classifying superfluous masking regions and moving objects.

This paper studies an adaptive high gain observer design for nonlinear systems which have lower triangular nonlinearity with Lipschitz coefficient, depending on the control input. Because the gain of the proposed observer is tuned automatically by a simple update law, our design approach doesn't need any information about the Lipschitz constant. Also, it is shown that under some assumptions, the dynamic gain of the proposed observer is bounded and its estimation error converges to zero asymptotically. Finally, a numerical example is given to verify the effectiveness of our design approach.

We propose a speaker adaptation method based on the probabilistic principal component analysis (PPCA) of acoustic models. We define a training matrix which is represented in a two-way array and decompose the training models by PPCA to construct bases. In the two-way array representation, each training model is represented as a matrix and the columns of each training matrix are treated as training vectors. We formulate the adaptation equation in the maximum a posteriori (MAP) framework using the bases and the prior.

A new stock index trend analysis approach is proposed in this paper, which is based on a newly developed signal decomposition approach - adaptive Fourier decomposition (AFD). AFD can effectively extract the signal's primary trend, which specifically suits the Dow Theory based technique analysis. The proposed approach integrates two different kinds of forecasting approaches, including the Dow theory the RBF neural network. Effectiveness of the proposed approach is assessed through comparison with the direct RBF neural network approach. The result is proved to be promising.

We consider a unified approach to the tracking analysis of adaptive filters with error and matrix data nonlinearities. Using energy-conservation arguments, we not only derive earlier results in a unified manner, but we also obtain new performance results for more general adaptive algorithms without requiring the restriction of the regression data to a particular distribution. Numerical simulations support the theoretical results.

There have been many matching pursuit algorithms (MPAs) which handle the sparse signal recovery problem, called compressed sensing (CS). In the MPAs, the correlation step makes a dominant computational complexity. In this paper, we propose a new fast correlation method for the MPA when we use partial Fourier sensing matrices and partial Hadamard sensing matrices which are widely used as the sensing matrix in CS. The proposed correlation method can be applied to almost all MPAs without causing any degradation of their recovery performance. Also, the proposed correlation method can reduce the computational complexity of the MPAs well even though there are restrictions depending on a used MPA and parameters.

We present the adaptation of the acoustic models of hidden Markov models (HMMs) to the target speaker and noise environment using bilinear models. Acoustic models trained from various speakers and noise conditions are decomposed to build the bases that capture the interaction between the two factors. The model for the target speaker and noise is represented as a product of bases and two weight vectors. In experiments using the AURORA4 corpus, the bilinear model outperforms the linear model.

A smoothing method for minimum phone error linear regression (MPELR) is proposed in this paper. We show that the objective function for minimum phone error (MPE) can be combined with a prior mean distribution. When the prior mean distribution is based on maximum likelihood (ML) estimates, the proposed method is the same as the previous smoothing technique for MPELR. Instead of ML estimates, maximum a posteriori (MAP) parameter estimate is used to define the mode of prior mean distribution to improve the performance of MPELR. Experiments on a large vocabulary speech recognition task show that the proposed method can obtain 8.4% relative reduction in word error rate when the amount of data is limited, while retaining the same asymptotic performance as conventional MPELR. When compared with discriminative maximum a posteriori linear regression (DMAPLR), the proposed method shows improvement except for the case of limited adaptation data for supervised adaptation.

Sensor noise prevents the exact measurement of output, which makes it difficult to guarantee the ultimate bound of the actual output and states, which is smaller than the sensor noise amplitude. Even worse, the time-varying delay in the input does not guarantee the boundedness of the actual output and states under sensor noise. In this letter, our considered system is a chain of integrators in which time-varying delay exists in the input and there is an additive form of sensor noise in the output measurement. To guarantee the arbitrarily small ultimate bound of the actual output and states, we newly propose an adaptive output feedback controller whose gain is tuned on-line. The merits of our control method over the existing results are clearly shown in the example.

Ultra-wide band (UWB) radar has a great advantage for range resolution, and is suitable for 3-dimensional (3-D) imaging sensor, such as for rescue robots or surveillance systems, where an accurate 3-dimensional measurement, impervious to optical environments, is indispensable. However, in indoor sensing situations, an available aperture size is severely limited by obstacles such as collapsed furniture or rubles. Thus, an estimated region of target image often becomes too small to identify whether it is a human body or other object. To address this issue, we previously proposed the image expansion method based on the ellipse extrapolation, where the fitting space is converted from real space to data space defined by range points to enhance the extrapolation accuracy. Although this method achieves an accurate image expansion for some cases, by exploiting the feature of the efficient imaging method as range points migration (RPM), there are still many cases, where it cannot maintain sufficient extrapolation accuracy because it only employs the single scattered component for imaging. For more accurate extrapolation, this paper extends the above image expansion method by exploiting double-scattered signals between the target and the wall in an indoor environment. The results from numerical simulation validate that the proposed method significantly expands the extrapolated region for multiple elliptical objects, compared with that obtained using only single scattered signal.

A switchable microwave reflector, reflection of which is actively controlled using diodes was proposed. Pin diodes have large resistance and capacitance without DC bias and small resistance and inductance with DC bias in microwave band. The reflector was designed by using the characteristics. In this paper, effects of a periodic structure on the reflector were verified with simulation and equivalent circuit model. A prototype reflector was able to switch between about $-20$ dB and $-0.1$ dB reflection coefficient at 2 GHz.

Negative Bias Temperature Instability (NBTI) is one of the serious concerns for long-term circuit performance degradation. NBTI degrades PMOS transistors under negative bias, whereas they recover once negative bias is removed. In this paper, we propose a mitigation method for NBTI-induced performance degradation that exploits the recovery property by shifting random input sequence through scan paths. With this method, we prevent consecutive stress that causes large degradation. Experimental results reveal that random scan-in vectors successfully mitigate NBTI and the path delay degradation is reduced by 71% in a test case when standby mode occupies 10% of total time. We also confirmed that 8-bit LFSR is capable of random number generation for this purpose with low area and power overhead.

Most images obtained with imaging sensors contain Bayer patterns and suffer from blurring caused by the lens. In order to convert a blurred Bayer-patterned image into a viewable image, demosaicing and deblurring are needed. These concepts have been major research areas in digital image processing for several decades. Despite their importance, their performance and efficiency are not satisfactory when considered independently. In this paper, we propose a joint deblurring and demosaicing method in which edge direction and edge strength are estimated in the Bayer domain and then edge adaptive deblurring and edge-oriented interpolation are performed simultaneously from the estimated edge information. Experimental results show that the proposed method produces better image quality than conventional algorithms in both objective and subjective terms.

Electromagnetic scattering problems of canonical 2D structures can be analyzed with a high degree of accuracy by using the point matching method with mode expansion. In this paper, we will extend our previous method to 3D electromagnetic scattering problems and investigate the radar cross section of spherical shells and the computational accuracy.

Acoustic feedback is a major complaint of hearing aid users. Adaptive filters are a common method for suppressing acoustic feedback in digital hearing aids. In this letter, we propose a new variable step-size algorithm for normalized least mean square and an affine projection algorithm to combine with a variable step-size affine projection algorithm and global speech absence probability in an adaptive filter. The computer simulation used to test the proposed algorithm results in a lower misalignment error than the comparison algorithm at a similar convergence rate. Therefore, the proposed algorithm suggests an effective solution for the feedback suppression system of digital hearing aids.

This letter proposes an adaptive base plane filtering algorithm for the inter-plane estimation of RGB images in HEVC RExt. Because most high-frequency components of RGB images have low inter-plane correlation, our proposed scheme adaptively removes the high-frequency components of the base plane in order to enhance the inter-plane estimation accuracy. The experimental results show that the proposed scheme provides average BD rate gains of 0.6%, 1.0%, and 1.2% in the G, B, and R planes, respectively, with slightly decreased complexity, as compared to the previous inter-plane filtering method.

Recognition errors caused by out-of-vocabulary (OOV) words lead critical problems when developing spoken language understanding systems based on automatic speech recognition technology. And automatic vocabulary adaptation is an essential technique to solve these problems. In this paper, we propose a novel and effective automatic vocabulary adaptation method. Our method selects OOV words from relevant documents using combined scores of semantic and acoustic similarities. Using this combined score that reflects both semantic and acoustic aspects, only necessary OOV words can be selected without registering redundant words. In addition, our method estimates probabilities of OOV words using semantic similarity and a class-based N-gram language model. These probabilities will be appropriate since they are estimated by considering both frequencies of OOV words in target speech data and the stable class N-gram probabilities. Experimental results show that our method improves OOV selection accuracy and recognition accuracy of newly registered words in comparison with conventional methods.

This study proposes a variable selection linear regression (VSLR) adaptation framework to improve the accuracy of automatic speech recognition (ASR) with only limited and unlabeled adaptation data. The proposed framework can be divided into three phases. The first phase prepares multiple variable subsets by applying a ranking filter to the original regression variable set. The second phase determines the best variable subset based on a pre-determined performance evaluation criterion and computes a linear regression (LR) mapping function based on the determined subset. The third phase performs adaptation in either model or feature spaces. The three phases can select the optimal components and remove redundancies in the LR mapping function effectively and thus enable VSLR to provide satisfactory adaptation performance even with a very limited number of adaptation statistics. We formulate model space VSLR and feature space VSLR by integrating the VS techniques into the conventional LR adaptation systems. Experimental results on the Aurora-4 task show that model space VSLR and feature space VSLR, respectively, outperform standard maximum likelihood linear regression (MLLR) and feature space MLLR (fMLLR) and their extensions, with notable word error rate (WER) reductions in a per-utterance unsupervised adaptation manner.

Polarimetric coherence strongly relates to the types and orientations of local scatterers. An optimization scheme is proposed to optimize the coherence between two polarimetric channels for polarimetric SAR (PolSAR) data. The coherence magnitude (correlation coefficient) is maximized by rotating a polarimetric coherence matrix in the rotation domain around the radar line of sight. L-band E-SAR and X-band Pi-SAR PolSAR data sets are used for demonstration and validation. The coherence of oriented manmade targets is significantly enhanced while that of forests remains relatively low. Therefore, the proposed technique can effectively discriminate these two land covers which are easily misinterpreted by the conventional model-based decomposition. Moreover, based on an optimized polarimetric coherence parameter and the total backscattered power, a simple manmade target extraction scheme is developed for application demonstration. This approach is applied with the Pi-SAR data. The experimental results validate the effectiveness of the proposed method.