This paper proposes a new quantization framework on activation function of deep neural networks (DNN). We implement fixed-point DNN by quantizing the activations into powers-of-two integers. The costly multiplication operations in using DNN can be replaced with low-cost bit-shifts to massively save computations. Thus, applying DNN-based speech recognition on embedded systems becomes much easier. Experiments show that the proposed method leads to no performance degradation.

With the surge of social media platform, users' profile information become treasure to enhance social network services. However, attributes information of most users are not complete, thus it is important to infer latent attributes of users. Contemporary attribute inference methods have a basic assumption that there are enough labeled data to train a model. However, in social media, it is very expensive and difficult to label a large amount of data. In this paper, we study the latent attribute inference problem with very small labeled data and propose the SRW-COND solution. In order to solve the difficulty of small labeled data, SRW-COND firstly extends labeled data with a simple but effective greedy algorithm. Then SRW-COND employs a supervised random walk process to effectively utilize the known attributes information and link structure of users. Experiments on two real datasets illustrate the effectiveness of SRW-COND.

The most commonly used scattering parameters (S parameters) are normalized to a real reference resistance, typically 50Ω. In some cases, the use of S parameters normalized to some complex reference impedance is essential or convenient. But there are different definitions of complex-referenced S parameters that are incompatible with each other and serve different purposes. To make matters worse, different simulators implement different ones and which ones are implemented is rarely properly documented. What are possible scenarios in which using the right one matters? This tutorial-style paper is meant as an informal and not overly technical exposition of some such confusing aspects of S parameters, for those who have a basic familiarity with the ordinary, real-referenced S parameters.

Sensor networks are often used to understand underlying phenomena that are reflected through sensing data. In real world applications, this understanding supports decision makers attempting to access a disaster area or monitor a certain event regularly and thus necessary actions can be triggered in response to the problems. Practitioners designing such systems must overcome difficulties due to the practical limitations of the data and the fidelity of a network condition. This paper explores the design of a network solution for the data acquisition domain with the goal of increasing the efficiency of data gathering efforts. An unmanned aerial vehicle (UAV) is introduced to address various real-world sensor network challenges such as limited resources, lack of real-time representative data, and mobility of a relay station. Towards this goal, we introduce a novel cooperative path selection framework to effectively collect data from multiple sensor sources. The framework consists of six main parts ranging from the system initialization to the UAV data acquisition. The UAV data acquisition is useful to increase situational awareness or used as inputs for data manipulation that support response efforts. We develop a system-based simulation that creates the representative sensor networks and uses the UAV for collecting data packets. Results using our proposed framework are analyzed and compared to existing approaches to show the efficiency of the scheme.

This study presents a two-stage spoken term detection (STD) method that uses the same STD engine twice and a support vector machine (SVM)-based classifier to verify detected terms from the STD engine's output. In a front-end process, the STD engine is used to pre-index target spoken documents from a keyword list built from an automatic speech recognition result. The STD result includes a set of keywords and their detection intervals (positions) in the spoken documents. For keywords having competitive intervals, we rank them based on the STD matching cost and select the one having the longest duration among competitive detections. The selected keywords are registered in the pre-index. They are then used to train an SVM-based classifier. In a query term search process, a query term is searched by the same STD engine, and the output candidates are verified by the SVM-based classifier. Our proposed two-stage STD method with pre-indexing was evaluated using the NTCIR-10 SpokenDoc-2 STD task and it drastically outperformed the traditional STD method based on dynamic time warping and a confusion network-based index.

A novel experimental design method based on a low-frequency active load-pull technique that includes harmonic tuning has been proposed for high-efficiency microwave power amplifiers. The intrinsic core component of a transistor with a maximum oscillation frequency of more than several tens of gigahertz can be approximately assumed as the nonlinear current source with no frequency dependence at an operation frequency of several gigahertz. In addition, the reactive parasitic elements in a transistor can be omitted at a frequency of much less than 1GHz. Therefore, the optimum impedance condition including harmonics for obtaining high efficiency in a nonlinear current source can be directly investigated based on a low-frequency active harmonic load-pull technique in the low-frequency region. The optimum load condition at the operation frequency for an external load circuit can be estimated by considering the properties of the reactive parasitic elements and the nonlinear current source. For an InGaAs/GaAs pHEMT, active harmonic load-pull considering up to the fifth-order harmonic frequency was experimentally carried out at the fundamental frequency of 20MHz. By using the estimated optimum impedance condition for an equivalent nonlinear current source, high-frequency amplifiers were designed and fabricated at the 1.9-GHz, 2.45-GHz, and 5.8-GHz bands. The fabricated amplifiers exhibited maximum drain efficiency values of 79%, 80%, and 74% at 1.9GHz, 2.47GHz, and 5.78GHz, respectively.

As an important extension of high-efficiency video coding (HEVC), screen content coding (SCC) includes various new coding modes, such as Intra Block Copy (IBC), Palette-based coding (Palette), and Adaptive Color Transform (ACT). These new tools have improved screen content encoding performance. This paper proposed a novel and fast algorithm by classifying Code Units (CUs) as text CUs or non-text CUs. For text CUs, the Intra mode was skipped in the compression process, whereas for non-text CUs, the IBC mode was skipped. The current CU depth range was then predicted according to its adjacent left CU depth level. Compared with the reference software HM16.7+SCM5.4, the proposed algorithm reduced encoding time by 23% on average and achieved an approximate 0.44% increase in Bjøntegaard delta bit rate and a negligible peak signal-to-noise ratio loss.

This paper proposes statistical analysis of phase-only correlation functions with phase-spectrum differences following wrapped distributions. We first assume phase-spectrum differences between two signals to be random variables following a linear distribution. Next, based on directional statistics, we convert the linear distribution into a wrapped distribution by wrapping the linear distribution around the circumference of the unit circle. Finally, we derive general expressions of the expectation and variance of the POC functions with phase-spectrum differences following wrapped distributions. We obtain exactly the same expressions between a linear distribution and its corresponding wrapped distribution.

Collaborative filtering (CF) has been widely used in recommender systems to generate personalized recommendations. However, recommender systems using CF are vulnerable to shilling attacks, in which attackers inject fake profiles to manipulate recommendation results. Thus, shilling attacks pose a threat to the credibility of recommender systems. Previous studies mainly derive features from characteristics of item ratings in user profiles to detect attackers, but the methods suffer from low accuracy when attackers adopt new rating patterns. To overcome this drawback, we derive features from properties of item popularity in user profiles, which are determined by users' different selecting patterns. This feature extraction method is based on the prior knowledge that attackers select items to rate with man-made rules while normal users do this according to their inner preferences. Then, machine learning classification approaches are exploited to make use of these features to detect and remove attackers. Experiment results on the MovieLens dataset and Amazon review dataset show that our proposed method improves detection performance. In addition, the results justify the practical value of features derived from selecting patterns.

This paper presents a weighted diversity combining technique for the cyclostationarity detection based spectrum sensing of orthogonal frequency division multiplexing signals in cognitive radio. In cognitive radio systems, secondary users must detect the desired signal in an extremely low signal-to-noise ratio (SNR) environment. In such an environment, multiple antenna techniques (space diversity) such as maximum ratio combining are not effective because the energy of the target signal is also extremely weak, and it is difficult to synchronize some received signals. The cyclic autocorrelation function (CAF) is used for traditional cyclostationarity detection based spectrum sensing. In the presented technique, the CAFs of the received signals are combined, while the received signals themselves are combined with general space diversity techniques. In this paper, the value of the CAF at peak and non-peak cyclic frequencies are computed, and we attempt to improve the sensing performance by using different weights for each CAF value. The results were compared with those from conventional methods and showed that the presented technique can improve the spectrum sensing performance.

Abstract The concept, state of the art, and future development directions of hybrid semiconductor integrated circuits (HySICs), which combine RF-CMOS ICs with compound semiconductor monolithic microwave integrated circuits (MMICs) are described in this paper, taking up recent wireless technologies as example applications. It is shown that ICs with superior function can be designed by mixing the optimal characteristics from the different semiconductors. To realize new semiconductor ICs, several component technologies for RF-HySIC are introduced in terms of chip/MMIC design, measurement, and breadboard model fabrication. A prototype RF-HySIC is described for the combination of a GaN Schottky barrier diode with a Si RF-IC matching network developed at 5.8GHz. A GaN diode structure, measurement and characterization of nonlinear devices, a GaN amplifier, and a GaAs MMIC are introduced as component technologies. In addition, the design for using an RF-CMOS matching network circuit with a size of 1.2mm × 2.3mm and room-temperature chip/wafer direct bonding under high-pressure conditions are explained. For advanced and autonomous ICs, HySIC and chip/MMIC topologies combined with a processor are proposed for application of HySIC to wireless sensor systems.

The introduction of deep neural networks (DNNs) leads to a significant improvement of the automatic speech recognition (ASR) performance. However, the whole ASR system remains sophisticated due to the dependent on the hidden Markov model (HMM). Recently, a new end-to-end ASR framework, which utilizes recurrent neural networks (RNNs) to directly model context-independent targets with connectionist temporal classification (CTC) objective function, is proposed and achieves comparable results with the hybrid HMM/DNN system. In this paper, we investigate per-dimensional learning rate methods, ADAGRAD and ADADELTA included, to improve the recognition of the end-to-end system, based on the fact that the blank symbol used in CTC technique dominates the output and these methods give frequent features small learning rates. Experiment results show that more than 4% relative reduction of word error rate (WER) as well as 5% absolute improvement of label accuracy on the training set are achieved when using ADADELTA, and fewer epochs of training are needed.

Antenna radiation patterns have side-lobes that add to ambiguity in the form of ghosting and object repetition in SAR images. An L-band 1.27GHz, 2×5 element proximity-coupled corner-truncated patch array antenna synthesized using the Dolph-Chebyshev method to reduce side-lobe levels is proposed. The designed antenna was sim-ulated, optimized, and fabricated for antenna performance parameter measurements. Antenna performance characteristics show good agree-ment with simulated results. A set of antennas were fabricated and then used together with a custom synthetic aperture radar system and SAR imaging performed on a point target in an anechoic chamber. Imaging results are also discussed in this paper showing improvement in image output. The antenna and its connected SAR systems developed in this work are different from most previous work in that this work is utilizing circular polarization as opposed to linear polarization.

This study proposes a two-pass spoken term detection (STD) method. The first pass uses a phoneme-based dynamic time warping (DTW)-based STD, and the second pass recomputes detection scores produced by the first pass using conditional random fields (CRF)-based triphone detectors. In the second-pass, we treat STD as a sequence labeling problem. We use CRF-based triphone detection models based on features generated from multiple types of phoneme-based transcriptions. The models train recognition error patterns such as phoneme-to-phoneme confusions in the CRF framework. Consequently, the models can detect a triphone comprising a query term with a detection probability. In the experimental evaluation of two types of test collections, the CRF-based approach worked well in the re-ranking process for the DTW-based detections. CRF-based re-ranking showed 2.1% and 2.0% absolute improvements in F-measure for each of the two test collections.

We propose a novel method for estimating acoustic scenes such as user activities, e.g., “cooking,” “vacuuming,” “watching TV,” or situations, e.g., “being on the bus,” “being in a park,” “meeting,” utilizing the information of acoustic events. There are some methods for estimating acoustic scenes that associate a combination of acoustic events with an acoustic scene. However, the existing methods cannot adequately express acoustic scenes, e.g., “cooking,” that have more than one subordinate category, e.g., “frying ingredients” or “plating food,” because they directly associate acoustic events with acoustic scenes. In this paper, we propose an acoustic scene estimation method based on a hierarchical probabilistic generative model of an acoustic event sequence taking into account the relation among acoustic scenes, their subordinate categories, and acoustic event sequences. In the proposed model, each acoustic scene is represented as a probability distribution over their unsupervised subordinate categories, called “acoustic sub-topics,” and each acoustic sub-topic is represented as a probability distribution over acoustic events. Acoustic scene estimation experiments with real-life sounds showed that the proposed method could correctly extract subordinate categories of acoustic scenes.

Traditional action recognition approaches use pre-defined rigid areas to process the space-time information, e.g. spatial pyramids, cuboids. However, most action categories happen in an unconstrained manner, that is, the same action in different videos can happen at different places. Thus we need a better video representation to deal with the space-time variations. In this paper, we introduce the idea of mining spatial temporal saliency. To better handle the uniqueness of each video, we use a space-time over-segmentation approach, e.g. supervoxel. We choose three different saliency measures that take not only the appearance cues, but also the motion cues into consideration. Furthermore, we design a category-specific mining process to find the discriminative power in each action category. Experiments on action recognition datasets such as UCF11 and HMDB51 show that the proposed spatial temporal saliency video representation can match or surpass some of the state-of-the-art alternatives in the task of action recognition.

Exploring a parsimonious model that is just enough to represent the temporal dependency of time serial signals such as audio or speech is a practical requirement for many signal processing applications. A well suited method for intuitively and efficiently representing magnitude spectra is to use convolutive non-negative matrix factorization (CNMF) to discover the temporal relationship among nearby frames. However, the model order selection problem in CNMF, i.e., the choice of the number of convolutive bases, has seldom been investigated ever. In this paper, we propose a novel Bayesian framework that can automatically learn the optimal model order through maximum a posteriori (MAP) estimation. The proposed method yields a parsimonious and low-rank approximation by removing the redundant bases iteratively. We conducted intuitive experiments to show that the proposed algorithm is very effective in automatically determining the correct model order.

In this letter, a method of wideband direction of arrival (DOA) estimation based on co-prime arrays with sub-Nyquist sampling is proposed. Previous works have employed co-prime arrays for wideband DOA estimation, which can increase the degrees of freedom (DOFs) in the spatial domain. However, they are all based on Nyquist sampling. Different from existing methods, we incorporate a sub-Nyquist sampling scheme called multicoset sampling for DOA estimation to relax hardware condition. Simulation results show the correctness and effectiveness.

Pair-wise testing is an effective test planning technique for finding interaction faults using a small set of test cases. Constraint elicitation is an important process in the pair-wise testing design since constraints determine the test space; however, the constraint elicitation process has not been well studied. It usually requires manual capturing and precise definition of constraints. In this paper, we propose a constraint elicitation process that helps combinatorial test design. Our elicitation process consists of two steps: parameter combination identification and value pair determination. We conduct experiments on some test models, and demonstrate that some extracted rules match constraints and others helps to define constraints.

The unconditionally stable (US) Laguerre-FDTD method has recently attracted significant attention for its high efficiency and accuracy in modeling fine structures. One of the most attractive characteristics of this method is its marching-on-in-order solution scheme. This paper presents Hermite-Rodriguez functions as another type of orthogonal basis to implement a new 2-D US solution scheme.