Stream fusion for multi-stream automatic speech recognition

Multi-stream automatic speech recognition (MS-ASR) has been confirmed to boost the recognition performance in noisy conditions. In this system, the generation and the fusion of the streams are the essential parts and need to be designed in such a way to reduce the effect of noise on the final decision. This paper shows how to improve the performance of the MS-ASR by targeting two questions; (1) How many streams are to be combined, and (2) how to combine them. First, we propose a novel approach based on stream reliability to select the number of streams to be fused. Second, a fusion method based on Parallel Hidden Markov Models is introduced. Applying the method on two datasets (TIMIT and RATS) with different noises, we show an improvement of MS-ASR.     

Input‐output triggered control using ‐stability over finite horizons

This paper investigates stability of nonlinear control systems under intermittent information. Following recent results in the literature, we replace the traditional periodic paradigm, where the up‐to‐date information is transmitted and control laws are executed in a periodic fashion, with the event‐triggered paradigm. Building on the small gain theorem, we develop input–output triggered control algorithms yielding stable closed‐loop systems. In other words, based on the currently available (but outdated) measurements of the outputs and external inputs of a plant, a mechanism triggering when to obtain new measurements and update the control inputs is provided. Depending on the noise in the environment, the developed algorithm yields stable, asymptotically stable, and ‐stable (with bias) closed‐loop systems. Control loops are modeled as interconnections of hybrid systems for which novel results on ‐stability are presented. The prediction of a triggering event is achieved by employing ‐gains over a finite horizon. By resorting to convex programming, a method to compute ‐gains over a finite horizon is devised. Finally, our approach is successfully applied to a trajectory tracking problem for unicycles. Copyright © 2014 John Wiley & Sons, Ltd.     

Neural Network and Classification Approach in Identifying Customer Behavior in the Banking Sector: A Case Study of an International Bank

The customer relationship focus for banks is in development of main competencies and strategies of building strong profitable customer relationships through considering and managing the customer impression, influence on the culture of the bank, satisfactory treatment, and assessment of valued relationship building. Artificial neural networks (ANNs) are used after data segmentation and classification, where the designed model register records into two class sets, that is, the training and testing sets. ANN predicts new customer behavior from previously observed customer behavior after executing the process of learning from existing data. This article proposes an ANN model, which is developed using a six‐step procedure. The back‐propagation algorithm is used to train the ANN by adjusting its weights to minimize the difference between the current ANN output and the desired output. An evaluation process is conducted to determine whether the ANN has learned how to perform. The training process is halted periodically, and its performance is tested until an acceptable result is obtained. The principles underlying detection software are grounded in classical statistical decision theory.     

Agustin de Betancourt's telegraph: Study and virtual reconstruction

The transmission of information has gone through various stages of evolution throughout its history. A stage before that of the electric telegraph was the so-called aerial/optical telegraph. It was developed towards the end of the 18th century and was in service until the middle of the 19th century. Chappe's system was widely used in France, and was the first to be in consistent use. However, a new and technologically superior system was developed soon afterwards which superseded it. Its inventor was Agustín de Betancourt, considered by some authors one of the founders of the Theory of Machines and Mechanisms, who, together with the distinguished clockmaker Breguet, presented it to the French Authorities in the turbulent decade of the 1790s.This article presents a historical review of this telegraph and analyzes its technical characteristics. It presents analytically, numerically and graphically some of the statements made about the telegraph, and corrects other subsequent observations. Lastly, a detailed reconstruction of the telegraph is made using different advanced CAD techniques, which provide an accurate static and dynamic view of each of its parts.     

Patchy solution of a Francis–Byrnes–Isidori partial differential equation

The solution to the nonlinear output regulation problem requires one to solve a first‐order partial differential equation, known as the Francis–Byrnes–Isidori equations. In this paper, we propose a method to compute approximate solutions to the Francis–Byrnes–Isidori equations when the zero dynamics of the plant are hyperbolic and the exosystem is two dimensional. With our method, we are able to produce approximations that converge uniformly to the true solution. Our method relies on the periodic nature of two‐dimensional analytic center manifolds. Copyright © 2012 John Wiley & Sons, Ltd.     

A hardware design of a massive-parallel, modular NN-based vector quantizer for real-time video coding

This report describes the design of a modular, massive-parallel, neural-network (NN)-based vector quantizer for real-time video coding. The NN is a self-organizing map (SOM) that works only in the training phase for codebook generation, only at the recall phase for real-time image coding, or in both phases for adaptive applications. The neural net can be learned using batch or adaptive training and is controlled by an inside circuit, finite-state machine-based hard controller. The SOM is described in VHDL and implemented on electrically (FPGA) and mask (standard-cell) programmable devices.     

Two proposals for the inclusion of directory information in the last-level private caches of glueless shared-memory multiprocessors

In glueless shared-memory multiprocessors where cache coherence is usually maintained using a directory-based protocol, the fast access to the on-chip components (caches and network router, among others) contrasts with the much slower main memory. Unfortunately, directory-based protocols need to obtain the sharing status of every memory block before coherence actions can be performed. This information has traditionally been stored in main memory, and therefore these cache coherence protocols are far from being optimal. In this work, we propose two alternative designs for the last-level private cache of glueless shared-memory multiprocessors: the lightweight directory and the SGluM cache. Our proposals completely remove directory information from main memory and store it in the home node’s L2 cache, thus reducing both the number of accesses to main memory and the directory memory overhead. The main characteristics of the lightweight directory are its simplicity and the significant improvement in the execution time for most applications. Its drawback, however, is that the performance of some particular applications could be degraded. On the other hand, the SGluM cache offers more modest improvements in execution time for all the applications by adding some extra structures that cope with the cases in which the lightweight directory fails.     

Bandwidth selection in kernel density estimation for interval-grouped data

When interval-grouped data are available, the classical Parzen–Rosenblatt kernel density estimator has to be modified to get a computable and useful approach in this context. The new nonparametric grouped data estimator needs of the choice of a smoothing parameter. In this paper, two different bandwidth selectors for this estimator are analyzed. A plug-in bandwidth selector is proposed and its relative rate of convergence obtained. Additionally, a bootstrap algorithm to select the bandwidth in this framework is designed. This method is easy to implement and does not require Monte Carlo. Both proposals are compared through simulations in different scenarios. It is observed that when the sample size is medium or large and grouping is not heavy, both bandwidth selection methods have a similar and good performance. However, when the sample size is large and under heavy grouping scenarios, the bootstrap bandwidth selector leads to better results.     

Mild Binding of Protein to C2N Monolayer Reveals Its Suitable Biocompatibility

The development of biocompatible nanomaterials for smart drug delivery and bioimaging has attracted great interest in recent years in biomedical fields. Here, the interaction between the recently reported nitrogenated graphene (C₂N) and a prototypical protein (villin headpiece HP35) utilizing atomistic molecular dynamics simulations is studied. The simulations reveal that HP35 can form a stable binding with the C₂N monolayer. Although the C₂N–HP35 attractive interactions are constantly preserved, the binding strength between C₂N and the protein is mild and does not cause significant distortion in the protein's structural integrity. This intrinsic biofriendly property of native C₂N is distinct from several widely studied nanomaterials, such as graphene, carbon nanotubes, and MoS₂, which can induce severe protein denaturation. Interestingly, once the protein is adsorbed onto C₂N surface, its transverse migration is highly restricted at the binding sites. This restriction is orchestrated by C₂N's periodic porous structure with negatively charged “holes,” where the basic residues—such as lysine—can form stable interactions, thus functioning as “anchor points” in confining the protein displacement. It is suggested that the mild, immobilized protein attraction and biofriendly aspects of C₂N would make it a prospective candidate in bio‐ and medical‐related applications.