Search

Consumers increasingly take online word-of-mouth as an important reference. Negative word-of-mouth spread rapidly on social media. How to effectively deal with the spreading of negative word-of-mouth on the platform is important to the enterprise. In order to predict the spread of negative word-of-mouth on the platform more accurately, we quantify user influence from four dimensions of user activity, user behavior, user authenticity, and user infection ability, and test the non-collinearity of these four dimensions to ensure the comprehensive and non-redundant evaluation. Then, combined with the Hidden Markov Model logic framework, an algorithm using user influence to predict the heat of negative word-of-mouth spreading was proposed. Meanwhile, we integrate both the static and dynamic information of microblog contents, directly quantify the popularity of negative word-of-mouth spreading on the platform as a benchmark, and select ten data sets from negative word-of-mouth transmission events to test the performance of the proposed algorithms. [ABSTRACT FROM AUTHOR]

Physician altruism is not only a key foundation of modern medical professionalism, but also a critical component in the theoretical health economics study. There is considerable interest in understanding the impacts of contemporary healthcare technology on physician altruism. In this paper, we seek to understand the dynamic influence of various motivating mechanisms on an online healthcare community (OHC). We characterize the dynamics of physician altruism on an OHC using a Hidden Markov Model (HMM) that allows the effects of motivating mechanisms to change across individuals and over time. We focus on the motivating mechanisms that give physicians social and economic returns. Our results paint a dynamic picture of physician altruism that is seldomly captured in previous studies. The findings are also essential to understanding the nature of physician altruism and the mechanism through which it can be influenced. Implications for the community owners and policy makers are discussed. [ABSTRACT FROM AUTHOR]

Player engagement determines his gaming activities and is of great importance to publishers. This paper investigates two dimensions of players' activities - playing volume and reward ads watching. To be specific, we propose a hidden Markov model that captures the player engagement as a hidden state and calibrate it on a detailed clickstream data from a mobile gaming company. Our findings reveal that (1) Cumulative challenge level can motivate medium engagement players to move to a higher state. (2) Cumulative playing time deters low engagement players from transiting a higher state, but keeps high engagement players from leaving. (3) There is a complementary effect between free coins obtained from mission completion and reward ads watching behavior among high engagement players, and we do not find evidence of substitution between in-game purchases and reward ads watching. [ABSTRACT FROM AUTHOR]

The Speech is most important & primary mode of Communication among of human being through which human being can communicate. The communication among the computer and human is called human computer interface. Speech is the important mode of interaction with computer. Today Speech Recognition Systems remains one of the most important research challenges to the researchers, e.g. noise, speaker variability, language variability, vocabulary size and domain. It require to much attention from the researchers to find out actual speech to convert in to digital or analog signals irrespective of disturbances. So that that speech can get converted in to digital form that is in the form of TEXT. This paper studies overall process of the same and gives an overview of Speech recognition System, how it works and what exactly it is. It also focus on advances technological perspective and appreciation of the fundamental progress of speech recognition. [ABSTRACT FROM AUTHOR]

Forecasting of future crude oil price in one of the most challenging problem in forecasting technique due to its haphazardness, non-linearity and complexity, many models are available to predict these non-linear and complexity. Recently, many researchers developed models based on Artificial Neural Network (ANN), Support Vector Machine (SVM), Fuzzy Logic (FL) and Moving Average (MA). In this paper, we developed a crude oil price forecasting model based on Hidden Markov Model due to its proven fittingness for modeling vigorous systems and pattern classification. We apply the HMM methodology to forecast the crude oil price from 1996 to 2015 using available past datasets from West Texas Intermediate. We used Matlab software in training and validating the data. The results obtained using Hidden Markov Model is inspiring because of the little error given. Lastly, we recommend development of fusion models by employing Hidden Markov Model with other lenient computing models. [ABSTRACT FROM AUTHOR]

The Speech is most important & primary mode of Communication among of human being through which human being can communicate. The communication among the computer and human is called human computer interface. Speech is the important mode of interaction with computer. Today Speech Recognition Systems remains one of the most important research challenges to the researchers, e.g. noise, speaker variability, language variability, vocabulary size and domain. It require to much attention from the researchers to find out actual speech to convert in to digital or analog signals irrespective of disturbances .So that that speech can get converted in to digital form that is in the form of TEXT. This paper studies overall process of the same and gives an overview of Speech recognition System, how it works and what exactly it is. It also focus on advances technological perspective and appreciation of the fundamental progress of speech recognition. [ABSTRACT FROM AUTHOR]

Sequence alignment is the basic method in sequence analysis, which is the process of composing or aligning two or more primary sequences so that the sequence similarity is apparent. One of the uses of this method is to predict the structure or function of an unknown protein by using a known protein information structure or function if the protein has the same sequence in database. Protein are macromolecules that make up more than half of the cell. Proteins are a chain of 20 amino acid combinations. Each type of protein has a unique number and sequence of amino acids. The method that can be applied for sequence alignment is the Genetic Algorithm, the other method is related to the Hidden Markov Model (HMM). The Hidden Markov Model (HMM) is a developmental form of the Markov Chain, which can be applied in cases that can not be directly observed. As Observed State (O) for sequence alignment is the sequence of amino acids in three categories: deletion, insertion and match. As for the Hidden State is the amino acid residue, which can determine the family protein corresponds to observation O. [ABSTRACT FROM AUTHOR]

Hijaiyah letters are the letters that arrange the words in Al Qur'an consisting of 28 letters. They symbolize the consonant sounds. On the other hand, the vowel sounds are symbolized by harokat/marks. Speech recognition system is a system used to process the sound signal to be data so that it can be recognized by computer. To build the system, some stages are needed i.e characteristics/feature extraction and classification. In this research, LPC and MFCC extraction method, K-Means Quantization vector and Hidden Markov Model classification are used. The data used are the 28 letters and 6 harakat with the total class of 168. After several are testing done, it can be concluded that the system can recognize the pronunciation pattern of marked hijaiyah letter very well in the training data with its highest accuracy of 96.1% using the feature of LPC extraction and 94% using the MFCC. Meanwhile, when testing system is used, the accuracy decreases up to 41%. [ABSTRACT FROM AUTHOR]

The paper presents an alternative vision of the problem of modeling the cyclic development and forecasting the type of the stage of the economic cycle. The suggested algorithm bases on the theory of hidden markov processes and includes two blocks: cyclic development analysis and cyclic development forecasting. Hidden markov model consists of two types of variables: unobservable hidden variable and variable, which is observable and measurable. The cycle stage indicator is offered as a hidden variable. The proposed algorithm was applied to analyze and forecast the cycle development of Ukrainian economy. [ABSTRACT FROM AUTHOR]

In the process of software development it is of high interest for a project manager to gain insights about the ongoing process and possible development trends at several points in time. Substantial factors influencing this process are, e.g., the constellation of the development team, the growth and complexity of the system, and the error-proneness of software entities. For this purpose we build an agent based simulation tool which predicts the future of a project under given circumstances, stored in parameters, which control the simulation process. We estimate these parameters with the help of software mining. Our work exposed the need for a more fine-grained model for the developer behavior. Due to this we create a learning model, which helps us to understand the contribution behavior of developers and, thereby, to determine simulation parameters close to reality. In this paper we present our agent based simulation model for software evolution and describe how methods from statistical learning and data mining serves us to estimate suitable simulation parameters. [ABSTRACT FROM AUTHOR]

Background: Dentists are subject to staying in static or awkward postures for long periods due to their highly concentrated work. Objectives: This study describes a realtime personalized biofeedback system developed for dental posture training with the use of vibrotactile biofeedback. Methods: The real-time personalized biofeedback system was an integrated solution that comprised of two components: 1) a wearable device that contained an accelerometer sensor for measuring the tilt angle of the body (input) and provided realtime vibrotactile biofeedback (output); and 2) software for data capturing, processing, and personalized biofeedback generation. The implementation of real-time personalized vibrotactile feedback was computed using Hidden Markov Models (HMMs). For the test case, we calculated the probability and log-likelihood of the test movements under the Work related Musculoskeletal Disorders (WMSD) and non- WMSD HMMs. The vibrotactile biofeedback was provided to the user via a wearable device for a WMSD-predicted case. In the system evaluation, a randomized crossover trial was conducted to compare dental posture measure using tilt angles of the upper back and muscle activities of those dental students that received vibrotactile biofeedback from the system with the control group against the dental students who received no feedback. Results: The participants who received feedback from the system had a lower tilt angle at 10th, 50th, and 90th percentiles of Backx and Backy, as well as muscular load, which were statistically different (p < 0.05) from those who received no feedback from the system. Conclusions: The results presented here demonstrate that a personalized biofeedback system for posture training in dental students is feasible and associated with quantitative improvements of the dental posture. [ABSTRACT FROM AUTHOR]

In recent years, many applications based on movement prediction have appeared very strongly. One of those is in the smooth handover process and resource reservation to improve Quality of Service (QoS). For effective productivity of these processes, time factor is very important. It needs exact information such as when and how long the user will go to a next location. Active Lezi algorithm and order-2 Markov model algorithm were appreciated in accuracy implementation, but the time factors were not significantly considered. In this paper, we propose a novel location prediction method constructed from a list of time and location history. A tree based on Lezi family of data compression is constructed and then, combined with Hidden Markov Model (HMM) for movement prediction driven by a sequence of time series. From our experiment, we used a part of dataset come from Dartmouth including information of 281 users. We achieved an accuracy rate of 85.52% location prediction. Moreover, the length of tree and the lost information were both reduced significantly. [ABSTRACT FROM PUBLISHER]

An increasing number of security incidents on industrial control systems drew a lot of concerns lately. Many attacks involved multiple attack vectors similar to internet attacks. However, CPS are more vulnerable to attacks. To evade detection, a hacker may apply multiple attack stages to gain the access to a control system. For example, he first employs a group of zombies (compromised machines) to identify the vulnerabilities of the target system and the findings would send back to the hacker through a communication channel. Once the correct access information is found by the zombies, the hacker could gain unauthorized access without violating any detection rules. The control system may be compromised by such multi-stage attacks and an appropriate defense mechanism is desired. In order to detect the sequence of such attack, this study correlates network information and system logs to find the stages of the attack. Finite state model, hidden Markov chain, is adopted to identify the multi-stage attacks and to prevent real damage. The results show that the proposed system can identify the multi-stage attacks in the early stage efficiently to prevent further damage in the networks. [ABSTRACT FROM PUBLISHER]

Most large systems such as HPC/cloud computing clusters and data centers are built from commercial off-the-shelf components. System logs are usually the main source of choice to gain insights into the system issues. Therefore, mining logs to diagnose anomalies has been an active research area. Due to the lack of organization and semantic consistency in commodity PC clusters' logs, what constitutes a fault or an error is subjective and thus building an automatic failure prediction model from log messages is hard. In this paper we sidestep the difficulty by asking a different question: Given the concomitant system log messages of a running job, can we predict the job's remaining time? We adopt Hidden Markov Model (HMM) coupled with frequency analysis to achieve this. Our HMM approach can predict 75% of jobs' remaining times with an error of less than 200 seconds. [ABSTRACT FROM PUBLISHER]

Integration testing is an expensive activity in software testing, especially for fault localization in complex systems. Model-based diagnosis (MBD) provides various benefits in terms of scalability and robustness. In this work, we propose a novel MBD approach for the automated fault localization in integration testing. Our method is based on Hidden Markov Model (HMM) which is an abstraction of system's component to simulate component's behaviour. The core of this method is a fault localization algorithm that gives out the set of suspect faulty components and a backward algorithm that calculates the matching degree between the HMM and the real system to evaluate the confidence degree of the localization conclusion. The proposed method is evaluated on a specific test bed and is applied to a simple helicopter control system case study. [ABSTRACT FROM PUBLISHER]

Nowadays HMMs are widely used for the online recognition of handwriting characters in different languages, and they are dominating as one of the standard choices for time series classification among research communities. Disadvantages of the systems using HMM are large training time as well as higher computational cost especially at higher number of states. Support Vector Machines is another statistical method based on the structural risk minimization principle and is capable of dealing with features of higher dimensionality. This paper presents the comparative performance of On-line Malayalam handwritten character recognition system using HMM and SVM. Time domain features such as input co-ordinates and the angular features are basically used to form the feature vector. A novel feature extraction method based on discrete wavelet transform is used for the system using SVM. The system gives maximum accuracy of 97.97% for SVM and 95.24% for HMM when tested on 1279 character samples, also the system using SVM require very much less training time compared to those using HMM. [ABSTRACT FROM PUBLISHER]

Speech Pre-processing is measured as major step in development of feature vector extraction for an efficient Automatic Speech Recognition (ASR) system. A novel approach for speech feature extraction is by applying the Mel-frequency cepstral co-efficient (MFCC) algorithm using Cubic-Log compression instead of Logarithmic compression in MFCC. In proposed MFCC, the frequency axis is initially warped to the mel-scale which is roughly below 2 kHz and logarithmic above this point. Triangular filter are equally spaced in the mel-scale are applied on the warped spectrum. The result of the filters are compressed using Cubic-Log function and cepstral co-efficient are computed by applying DCT to obtain minimum MFCC feature vector for spoken words. These feature vectors are given as input to classification and Recognition phase. The system is trained and tested by generating MFCC feature vector for 600 isolated words, 256 connected words and 150 sentences in clear and noisy environment. Experiment results shows that with minimum MFCC feature vector is enough for speech recognition system to achieve high recognition rate and its performance is measured based on Mean Square Error (MSE) rate. [ABSTRACT FROM PUBLISHER]

We have developed the 30K word real time continuous speech recognition based on a context dependent Hidden Markov Model (HMM). Here we are using a 30K word language model instead of previously using 20K[15] word speech recognition. It has opened new opportunities for speech recognition innovations. In 20K [15]word speech recognition has been designed with limited vocabulary .i.e., 800 words[9] but in this 30K word language model to be designed by using the high level vocabulary. In this system contains two parts. One is training and second is testing. First different input speech signals will be stored in training kit. Second will give different speech signals for testing, after comparing with training kit it will display final output. Gaussian Mixture Models (GMMs)[3] are used to represent the state of output probability of HMMs. [ABSTRACT FROM PUBLISHER]

The deployment of smart meters has made available high-frequency (minutes as opposed to monthly) measurements of electricity usage at individual households. Converting these measurements to knowledge that can improve energy efficiency in the residential sector is critical to attract further smart grid investments and engage electricity consumers in the path towards reducing global carbon footprint. The goal of the reported research is to use smart meter measurement data to identify heating and cooling usage levels for a home. This is important to cost effectively design consumer energy services such as energy audit and demand response targeted towards improving an individual household's heating usage efficiency. We present a machine learning approach akin to Non-Intrusive Load Monitoring (NILM) to disaggregate heating usage from measurements of a household's total electricity usage. We use as input 15-minute interval meter data and hourly outdoor temperature measurements. Our approach does not require a manual set-up procedure at each house. The method uses a Hidden Markov Model to capture the dependence of heating usage on outdoor temperature. Compared to existing methods based on linear regression, the proposed method provides details on heating usage patterns and is more flexible to incorporate other system specific information. Preliminary results based on synthetic and real-world usage data demonstrate the feasibility of the proposed approach. [ABSTRACT FROM PUBLISHER]

The Digital Sand Model is a tactical exercise without troops, which gives a bird's view of the terrain and allows the users to work on the battle planning and management. Interaction plays a substantial role in such training setups. In this paper we designed and developed an intuitive and "hands-free" human computer interaction for the DSM. The captured image is pre-processed for enhancing the hand object in the image. Specific features need to be extracted for fast and accurate recognition of the hand gesture. Two sets of feature vectors are explored in this paper, namely, Hu moments and hand contour. These features are classified using two different classifiers, namely, SVM and HMM and their performances are compared and analyzed. [ABSTRACT FROM PUBLISHER]

Speech recognition system is the method of implementing protocols and codes to enable a machine to take sound waves in the form of input, interpret it and formulate the desired output. In this paper, a method has been proposed to build up a speech recognition system which could interpret in any language (tested with Hindi and Bengali language) and implement it on Windows 7 platform which could be speaker independent i.e. recognition of voice irrespective of any random speaker. [ABSTRACT FROM PUBLISHER]

In this paper, a Hidden Markov Model (HMM) approach is introduced for recognizing human behaviors. Two main points are discussed in this approach: first is the application of HMM to the recognition system of human behaviors, and second is the effectiveness comparison of K-means and fuzzy C-means clustering algorithms. Three sample human behaviors are defined and the corresponded 3D data are collected using the Microsoft Kinect sensor (3D sensor). During these processing, we discuss the difference of k-means and fuzzy c-means clustering algorithms, and also we can see the results impacted by different clustering algorithms, the effectiveness of both clustering methods is shown through demonstrating the performance of our recognition system with HMM. [ABSTRACT FROM PUBLISHER]

We show that noise can speed training in hidden Markov models (HMMs). The new Noisy Expectation-Maximization (NEM) algorithm shows how to inject noise when learning the maximum-likelihood estimate of the HMM parameters because the underlying Baum-Welch training algorithm is a special case of the Expectation-Maximization (EM) algorithm. The NEM theorem gives a sufficient condition for such an average noise boost. The condition is a simple quadratic constraint on the noise when the HMM uses a Gaussian mixture model at each state. Simulations show that a noisy HMM converges faster than a noiseless HMM on the TIMIT data set. [ABSTRACT FROM PUBLISHER]

Current era is to make the interaction between humans and their artificial partners (Computers) and make communication easier and more reliable. One of the actual tasks is the use of vocal interaction. Speech recognition may be improved by visual information of human face. In literature, the lip shape and its movement are referred to as lip reading. Lip reading computing plays a vital role in automatic speech recognition and is an important step towards accurate and robust speech recognition. In this paper we consider our approach towards automatic lip reading in detail by using Active Appearance Model (AAM) and Hidden Markov Model (HMM). Appearance-based methods consider the raw image for feature extraction. AAM uses for detection of speaker's face features and feature points from faces automatically by appearance parameters of a speaker's lips. Hidden Markov Model (HMM) approach is used for Lip recognition. In this paper, we will describe the visual features based on shape and appearance descriptions and detailed information about the automatic lip reading system. Individual characteristics and efficiency of both approaches towards the Lip reading is also discuss. The efficiency of both approaches (AAM and HMM) will be evaluated individually by characteristics and their performances are then compared and discuss. In this paper we will show the recognition process of common visual features (i.e. selection of lip and extraction of movement) for an improved lip reading. [ABSTRACT FROM PUBLISHER]

The basic goal of developing the Phonetic Engine (PE) is to determine the sequence of basic sound units such as phones present in the spoken utterance. In this work, we are focusing on developing PEs for Indian languages namely Bengali and Oriya. This framework of developing PEs can be extended to any Indian languages. Two separate PEs for decoding the spoken utterances of Bengali and Oriya languages are developed. For developing the PEs, we have used read speech corpus. In this study, we have used 35 phones for developing PE for Bengali language and 32 phones for developing PE for Oriya language. We have developed PE by using Hidden Markov Models (HMMs) and FeedForward Neural Networks (FFNNs). Mel-frequency Cepstral Coefficients are used as features for building the models. In Speaker Dependent (SD) case, obtained accuracies of Bengali PE using HMMs and FFNNs are 41.65 and 53.87 percentages respectively for Speaker Dependent (SD) case. Likewise for Oriya PE, accuracies obtained using HMMs and FFNNs are 46.18 and 59.88 percentages respectively for SD case. [ABSTRACT FROM PUBLISHER]

This paper describes how to generate prosody of Indonesian sentences text to speech synthesis with a small corpus using hidden Markov model (HMM). Corpus contains PoS types and phrase categories types that represent text sentence syntax, as well as the type of pitch curve that represents the pattern of speech intonation. Learning process is carried out to obtain the probability of transition and emissions, where the hidden state input is a type of pitch curve pattern, and the state observation is the type of PoS and the type of syntax phrase categories. Text input sentences are syntactically parsed by chunking parser. HMM to predict the type of pitch curve patterns based on the output of chunking parser. Prosody generation is done by a certain rule that processes text input sentence, PoS type, phrase syntax categories type, and pitch curve pattern predict. Prosody generation using diphone concatenation with Indonesian diphone database. [ABSTRACT FROM PUBLISHER]

User mobility can cause severe degradation of QoS, which can be counteracted by performing handovers at the right moment. However, to perform a robust handover is a challenging task. In this paper, we study the handover problem for mobile wireless networks in the office environment, where the users' mobility is predictable. We propose a new mechanism using the Hidden Markov Model (HMM) to predict the Received Signal Strength (RSS) values. Based on the predicted values, each mobile user can perform the handover decision on his own using his preferred decision making algorithm. Extensive simulations were carried out for an office environment to evaluate the performance of the prediction handover scheme. The results showed that our HMM model can predict the RSS values accurately. The knowledge obtained from prediction is considered in the handover decision making algorithm, which can result in a seamless handover. Moreover, with the knowledge about the future RSS, mobile users can avoid multiple handovers in a short time known as “Ping-Pong” effect, which in turn reduces the signaling overheads on the network. [ABSTRACT FROM PUBLISHER]

We propose to simulate data series of human activities, in order to provide benchmarking for human activity recognition algorithms. Within the French project “AILISA” we recorded 1492 days of data of activity collected with presence sensors in our experimental Health Smart Homes. We built a mathematical model on the data series, based on “Hidden Markov Models” (HMM) and Urn of Polya. The model was then played on a computer to produce simulated data series with flexibility to adjust the parameters in various scenarios. We tested several methods to measure the similarity between our real and simulated data and obtained better results when using the surface correlation. [ABSTRACT FROM PUBLISHER]

In received signal strength fingerprints based indoor localization systems, the radio map built by labeled wireless fingerprints is easily outdated over time, while re-calibrating the overall radio map is time consuming. To avoid the tedious task, we propose to employ manifold alignment to label the current radio map from outdated radio map, with the constraint of the Hidden Markov Model trained by trajectories of the received signal strength readings. Manifold alignment can align the low-dimensional manifold structures of two different data sets and transfer knowledge across them. Transition matrix generated by Hidden Markov Model is used to constrain the alignment of manifolds. The proposed algorithms are tested in a real world ZigBee environment. Experiment results show that our method outperforms state-of-the-art transfer learning algorithms. [ABSTRACT FROM PUBLISHER]

The ineffectiveness of information retrieval systems is mostly caused by the inaccurate query formed by a few keywords that reflect actual user information need. One well known technique to overcome this limitation is Automatic Query Expansion (AQE), whereby the user's original query is improved by adding new features with a related meaning. It has long been accepted that capturing term associations is a vital part of information retrieval. It is therefore mainly to consider whether many sources of support may be combined to forecast term relations more precisely. This is mainly significant when frustrating to predict the probability of relevance of a set of terms given a query, which may involve both lexical and semantic relations between the terms. This paper presents a approach to expand the user query using three level domain model such as conceptual level(underlying Domain knowledge), linguistic level(term vocabulary based on Wordnet), stochastic model ME-HMM2 which combines (HMM (Hidden Markov Model and Maximum Entropy(ME) models) stores the mapping between such levels, taking into account the linguistic context of words. [ABSTRACT FROM PUBLISHER]

Video face recognition has become one of the most active research direction in the field of face recognition. In this paper, the influence of the illumination and the posture was analyzed and the corresponding solution was designed. Through the combined face recognition algorithm of global features and local features based on the wavelet analysis, as well as the Hidden Markov model using as the post-processing, the conception of optimizing recognition was set up for construction. The feasibility of the model was proved by the partial experiments, which provide the beneficial reference for intelligent monitoring technology. [ABSTRACT FROM PUBLISHER]

Nowadays, there are many fall detection systems based on intelligent video analysis. However, these systems are still facing many challenges such as lighting changes, long-term scene changes or added static background objects in new scene, etc. In this paper, adaptive background Gaussian mixture model (GMM) has been applied for moving object segmentation. An ellipse shape has been built from the segmented object for body modeling. Five features are extracted from this ellipse model and fed into two Hidden Markov Models (HMM) to classify fall and normal activities. We apply our proposed approach to challenging data sets recorded in different conditions. The qualitative results demonstrate that the combination of the adaptive GMM-based object segmentation and HMM certainly improves recognition accuracy under different scenarios. [ABSTRACT FROM PUBLISHER]

LSP has many advantages for speech representation, especially correlates well to spectrum formants as long as the LSP parameters are strictly ordered and bounded. This ordering property cannot be guaranteed during HMM-based speech synthesis when LSP is adopted as the spectrum feature, because diagonal covariance is utilized and correlation between LSP dimensions is ignored, with the result that unstable issue will be caused in synthesized speech. In this paper, we modify the parameter generation criterion to preserve ordering property of generated LSPs, by considering not only the likelihoods for HMM and GV maximized in conventional method but also a mis-orderings penalty. Experimental results show that the proposed method can alleviate the mis-orderings significantly and achieve high quality synthesizing performance when the penalty weight is selected appropriately. [ABSTRACT FROM PUBLISHER]

In this paper, 1-D Local binary patterns (LBP) are proposed to be used in speech signal segmentation and voice activation detection (VAD)and combined with hidden Markov model (HMM) for advanced speech recognition. Speech is firstly de-noised by Adaptive Empirical Model Decomposition (AEMD), and then processed using LBP based VAD. The short-time energy of the speech activity detected from the VAD is finally smoothed and used as the input of the HMM recognition process. The enhanced performance of the proposed system for speech recognition is compared with other VAD techniques at different SNRs ranging from 15 dB to a robust noisy condition at −5 dB. [ABSTRACT FROM PUBLISHER]

This paper describes a new hierarchical temporal Bayesian model and a Markov chain Monte Carlo (MCMC) algorithm for gene factor analysis. Each data sample is decomposed as a linear combination of characteristic gene signatures (also called factors) with appropriate proportions, or factor scores, following a linear mixing model (LMM). The particularity of the proposed algorithm is that the LMM model is combined with a hidden Markov model (HMM) to take into account temporal dependencies between the samples. The proposed HMM structure is motivated by the behavior of host molecular response following exposure to an infectious agent. The complexity of the posterior distribution resulting from the proposed HMM is alleviated by using a hybrid Gibbs sampler that generates samples distributed according to this distribution. These samples are then used to approximate the standard Bayesian estimators of the unknown parameters. The performance of the proposed method is illustrated by simulations conducted on synthetic data and on a real public dataset. [ABSTRACT FROM PUBLISHER]

The model of a router interface with a buffer limited to a fixed number of packets (regardless their lengths) is discussed. The interface is described as a finite FIFO queuing system with a Markovian Arrival Process (MAP) and semi-Markov (SM) service times. The new analytical results for the loss ratio, the local loss intensity or the total number of losses are developed. All results are suitable either for a transient or a stationary analysis and it is possible to extend them beyond the loss process. [ABSTRACT FROM PUBLISHER]

Recently, mobile communications need to benefit a good level of Quality of Service (QoS), since communications guarantees are mandatory during active flows. Passive resources are used to ensure service continuity when mobile hosts are moving among different coverage cells. In this work the attention is focused on wireless services in cellular networks, where the hand-over effects need to be mitigated, through an appropriate reservation policy. The whole considered system is modeled through a distributed set of Hidden Markov Chains (HMC) and the related theory is used to design a mobility predictor, as the main component of the proposed idea, which does not depend on the considered transmission technology (GSM, UMTS, WLAN, etc.), mobility model or vehicular scenario (urban, suburban, etc.). MRSVP has been used in order to realize the active/passive bandwidth reservation in the considered network topology and many simulation campaigns have been carried out in order to estimate the correctness of the proposed algorithm, also in terms of CDP and CBP. [ABSTRACT FROM PUBLISHER]

Internet is an infinite information repository that also contains harmful contents like; pornography, violence, and hate messages. It is very important to obstruct these kinds of contents from underage children not to adversely affect their development. Today, there are many commercial software products developed for this purpose. But the filtering capabilities of these commercial software are limited to text based and image based contents. Different techniques must be used to filter video based contents. This article describes an agent-based system which is developed for the detection of videos containing pornographic contents. Videos on the Internet can be divided into six groups as; anime, commercial, music, sitcom, sports, and porn related. The proposed system uses the Hidden Markov Model (HMM) based classification technique to classify the videos into these predefined categories with intelligent agents. Color features are extracted from each video frame and used as observation sequences in HMM for classification. According to the classification results, the videos, which are closely related to the category of sex and pornography, are filtered to the underage users. The test results obtained indicate that the classification has been satisfactory. [ABSTRACT FROM PUBLISHER]

The problem of designing a system for understanding user intention in human-computer dialogue is investigated in the article. Two different approaches have been suggested for solution of this problem: Hidden Markov and Neuro-Fuzzy models. Concrete object was selected for process of understanding user intention. The methods, which combines the using of hybrid neuro-fuzzy and hidden Markov models is offered. Mathematical algorithms and software of the system have been developed for automatically defining user intention in human-computer dialogue. [ABSTRACT FROM PUBLISHER]

This paper presents to apply Automatic Speech Recognition (ASR) algorithm to the meal service robot so that the user can be easier to order the meal and increase interaction between the robot and human. The Mel-frequency Cepstral coefficients (MFCC) is used to scratch the feature parameters, and Hidden Markov Model (HMM) is applied as the recognition speech model via Spce3200. It is based on the embedded HMM for speech recognition and thus reducing the size and power in a less computational time. There are 25 sets of Chinese speech data are trained and tested using a sentence and multi sentences under different environments. The highest recognition rate in a sentence has confirmed up to 95%, and multi-sentences can reach 85%. Accordingly, the practical results have indicated its relevant reliability. [ABSTRACT FROM PUBLISHER]

The design processes and methods of PHM system based on HMM have been investigated. HMM has some advantage in terms of dealing with small sample size and high discerning accuracy. The rationality of sensors set which is based on the hidden Markov model has been evaluated from quantitative point of view. Then the evaluating method of different sensor sets based on HMM has been put forward. At last, the effectiveness of process and method has been validated by using the data of gear case. [ABSTRACT FROM PUBLISHER]

Speaker-specific characteristics play an important role on the performance of Bangla (widely used as Bengali) automatic speech recognition (ASR). Gender factor shows adverse effect in the classifier while recognizing a speech by an opposite gender, such as, training a classifier by male but testing is done by female or vice-versa. To obtain a robust ASR system in practice it is necessary to invent a system that incorporates gender independent effect for particular gender. In this paper, we have proposed a Gender-Independent technique for ASR that focused on a gender factor. The proposed method trains the classifier with the both types of gender, male and female, and evaluates the classifier for the male and female. For the experiments, we have designed a medium size Bangla (widely known as Bengali) speech corpus for both the male and female. The proposed system has showed a significant improvement of word correct rates, word accuracies and sentence correct rates in comparison with the method that suffers from gender effects. Moreover, it requires a fewer mixture component in hidden Markov model (HMMs) and hence, computation time. [ABSTRACT FROM PUBLISHER]

A reasonably large database of online handwritten Bangla characters has been developed. Such a handwritten character sample is composed of one or more strokes. Seventy five such stroke classes have been identified on the basis of the varying handwriting styles present in the character database. Each character sample is a sequence of strokes emanating from these stroke classes. Another database of handwritten Bangla strokes has been developed from the character database. This is the first such database for Bangla script. Certain stroke level features are defined on the basis of certain extremum points which represent the stroke shape reasonably well. The proposed character classification method is a two-stage approach. First, a probability distribution is estimated for each stroke class using the stroke features and then an HMM based character classifier is designed using each stroke class as a state. The parameters of both the stroke class distributions and the character class HMMs are estimated on the basis of the training set having 29,951 character samples. The character level recognition accuracy obtained by the proposed method on the test set having 8,616 samples, is 91.85%. [ABSTRACT FROM PUBLISHER]

In this paper, a new offline Arabic handwriting recognition system is presented. The Douglas-Peucker algorithm is applied on the skeletonized parts of the offline images to convert it into piecewise linear curves that are used for efficient detection of diacritics, noise segments, and the baseline. A hidden Markov model (HMM)-based system is used with features extracted from the image before and after removing the diacritics. A reliable method of lexicon ranking and reduction based on the information of the image's diacritics, number of piece of Arabic words (PAWs), and dimensions information is used. The proposed system has been tested using the IFN/ENIT database and has achieved promising recognition rates. [ABSTRACT FROM PUBLISHER]

We present a statistical hypothesis testing method for handwritten word segmentation algorithms. Our proposed method can be used along with any word segmentation algorithm in order to detect over-segmented or under-segmented errors or to adapt the word segmentation algorithm to new data in an unsupervised manner. The main idea behind the proposed approach is to learn the geometrical distribution of words within a sentence using a Markov chain or a Hidden Markov Model (HMM). In the former, we assume all the necessary information is observable, where in the latter, we assume the minimum observable variables are the bounding boxes of the words, and the hidden variables are the part of speech information. Our experimental results on a benchmark database show that not only we can achieve a lower over-segmentation and under-segmentation error rate, but also a higher correct segmentation rate as a result of the proposed hypothesis testing. [ABSTRACT FROM PUBLISHER]

This paper presents the reconfigurable architecture and implementation of HMM-based decoder module in speech recognizer on FPGA. The architecture is done with different parameters of speech recognition system that could be easily reconfigurable. The design and implementation on FPGA have been verified with utterances of 800 test speeches. The implementation results, the recognition accuracy results of up to 98% are also presented. [ABSTRACT FROM PUBLISHER]

OFDM systems popularly being used in current high data rate wireless access networks and international standards are practically implemented using FEC and interleaving schemes whose performance evaluation and analysis can greatly benefit from Discrete Channel Models developed for the underlying channel. In this paper the order i.e., minimum number of states required for the Hidden Markov discrete channel modelling of a typical OFDM system has been estimated by extensive simulation. The input sample period (Ts) of the signal on which the channel acts, is seen to play a vital role in the HMMs for the OFDM systems. The models which have been completely derived for 4 different values of input sample period strongly indicate the need for specific evaluation of HMM channel models for different wireless standards. The Baum-Welch Algorithms (BWA) is used to train the HMM and to generate the HMM based error patterns which are compared with the original in terms of Autocorrelation functions and distribution probabilities of error intervals. [ABSTRACT FROM PUBLISHER]