HMM Model Research Articles

Research on Speech Recognition Robot Based on Embedded System and DSP

This paper has a discussion and research on the design and realization of the speech recognition robot based on embedded system and DSP. The solution of embedded system and DSP has made the performance, cost, reconfigurable ability and extensible ability of the system to a high level. And the system has adapted the modified MFCC method to extract the feature of the speech and used the HMM model based on K segmented equalizing value to do the speech study and recognition. It has improved the transplanting and real-time ability of the arithmetic.

A deep HMM model for multiple keywords spotting in handwritten documents

In this paper, we propose a query by string word spotting system able to extract arbitrary keywords in handwritten documents, taking both segmentation and recognition decisions at the line level. The system relies on the combination of a HMM line model made of keyword and non-keyword (filler) models, with a deep neural network that estimates the state-dependent observation probabilities. Experiments are carried out on RIMES database, an unconstrained handwritten document database that is used for benchmarking different handwriting recognition tasks. The obtained results show the superiority of the proposed framework over the classical GMM–HMM and standard HMM hybrid architectures.

Incorporating Intra-Query Term Dependencies in an Aspect Query Language Model

Query language modeling based on relevance feedback has been widely applied to improve the effectiveness of information retrieval. However, intra-query term dependencies i.e., the dependencies between different query terms and term combinations have not yet been sufficiently addressed in the existing approaches. This article aims to investigate this issue within a comprehensive framework, namely the Aspect Query Language Model AM. We propose to extend the AM with a hidden Markov model HMM structure to incorporate the intra-query term dependencies and learn the structure of a novel aspect HMM AHMM for query language modeling. In the proposed AHMM, the combinations of query terms are viewed as latent variables representing query aspects. They further form an ergodic HMM, where the dependencies between latent variables nodes are modeled as the transitional probabilities. The segmented chunks from the feedback documents are considered as observables of the HMM. Then the AHMM structure is optimized by the HMM, which can estimate the prior of the latent variables and the probability distribution of the observed chunks. Our extensive experiments on three large-scale text retrieval conference TREC collections have shown that our method not only significantly outperforms a number of strong baselines in terms of both effectiveness and robustness but also achieves better results than the AM and another state-of-the-art approach, namely the latent concept expansion model. © 2014Wiley Periodicals, Inc.

Automatic Target Detection Method for High-Resolution Synthetic Aperture Radar Images

Automatic target detection is of great importance in high-resolution synthetic aperture radar (SAR) images processing. In this paper, we proposed a hybrid HMM-TSVM model to detect targets in SAR images. Our proposed SAR image target detection system is made up of three steps. In this first step, the testing/training SAR images are pre-processed, and image visual features are extracted through 2DPCA, which is an improved version of principal component analysis. In the second step, we use the HMM model to construct the training sequence from training image dataset. Particularly, the relationships between the image in the azimuth and feature vectors are used to generate the feature sequences and training sequences for hidden Markov model. Furthermore, the feature sequences and training sequences are extracted from feature vectors with similar target images in an azimuth. In the third step, targets are detected from testing SAR images by TSVM classifier based on the training sequence by exchanging the labels of pair of different unlabeled samples to solve an objective function. Experimental results demonstrate the effectiveness of the proposed algorithm

Review of Stochastic POS Tagging Techniques used in Bengali

this paper, we describe different stochastic methods or techniques used for POS tagging of Bengali language. We have shown a generalized stochastic model for POS tagging in Bengali. We reviewed kinds of corpus and number of tags used for tagging methods. In the study it is found that as many as 45 useful tags existed in the literature. There are four useful corpus found in the study. As Bengali is a morphologically rich language we outlined a feature list that could be used with different training algorithms. We found that a hybrid HMM model used with a morphological analyzer work best in Bengali with an accuracy of 96.3%.

Combining Structure and Parameter Adaptation of HMMs for Printed Text Recognition.

We present two algorithms that extend existing HMM parameter adaptation algorithms (MAP and MLLR) by adapting the HMM structure. This improvement relies on a smart combination of MAP and MLLR with a structure optimization procedure. Our algorithms are semi-supervised: to adapt a given HMM model on new data, they require little labeled data for parameter adaptation and a moderate amount of unlabeled data to estimate the criteria used for HMM structure optimization. Structure optimization is based on state splitting and state merging operations and proceeds so as to optimize either the likelihood or a heuristic criterion. Our algorithms are successfully applied to the recognition of printed characters by adapting the HMM character models of a polyfont printed text recognizer to new fonts. Our experiments involve a total of 1,120,000 real and 3,100,000 synthetic character images and concern a set of 89 HMM models. A comparison of our results with those of state-of-the-art adaptation algorithms (MAP and MLLR) shows a significant increase in the accuracy of character recognition.

RETRACTED: Hybrid Hidden Markov Models and Neural Networks Based on Face Recognition

Retracted paper: In this paper, the basic principles of HMM, HMM studied three major issues need to be addressed as well as overflow problems in the practical application of how to solve the HMM. Because artificial neural network (ANN) with anti-noise, adaptive, learning ability, recognition speed, etc., taking into account the characteristics of the common features of speech recognition and pattern recognition and artificial neural networks have, this article will get a mixed combination of HMM in ANN model, using ANN to make up for some deficiencies of HMM. Experiments show that the hybrid model recognition rate than the HMM model increased by 4%, but the algorithm still has many defects to be resolved.

HMM modeling for OFDM–BER performance

OFDM based systems widely adopted in several wired and wireless standards can be analyzed efficiently with reduced cost, time and computation by characterizing the realistic physical layer with a Finite State Markov Channel. In this paper, accurate HMMs have been derived using BWA for a wide range of channel SNR conditions and different OFDM system parameters such as number of subcarriers, FFT size and number of symbols. OFDM with pilot based channel estimation errors have also been accurately parameterized for the symbol error rate. All estimated error statistics have been validated by comparison in terms of log likelihood, average error probability, autocorrelation function and the distributions of error or error free interval. Sequences from higher order modulation in OFDM systems can be reproduced with relatively fewer HMM states. The reliability of the model is confirmed by an excellent match to those of the underlying OFDM based physical layer. Accuracy of the ACF comparisons with number of HMM states has also been quantified.

Effective Fault Diagnosis Method for Composite Web Services Based on Improved HMM Model

Effective Fault Diagnosis Method for Composite Web Services Based on Improved HMM Model

Field Information Acquisition System Research Based on Offline Speech Recognition

In the mobile Internet era, the use of intelligent mobile terminal equipment will become more and more popular, especially in terms of using mobile phones for information collection in outdoor environments. Speech recognition technology is a very friendly method of interpersonal interaction, and using it for data acquisition will greatly improve collection efficiency. In this study, based on the theory of the HMM model, we build a wild information acquisition system which is based on offline continuous speech recognition through open source tools such as HTK and Sphinx. The system is able to efficiently solve some difficult problems, such as bad signals, inability to recognize rarely used terminology, and low speed of recognition.

An Enhanced Approach for Face Recognition of Newborns using HMM and SVD Coefficients

ABSTRACT The problems of newborn’s abduction, mixing, swapping, etc. are increasing day-by-day. This problem has now reached a global aspect, as the consequences have now become critical. The researches performed to meet the challenge are very few. The growing problem motivated a need of sustainable system that can assist the hospital authorities and even the parents to keep a track of any newborn baby and his parents. The concept of biometric recognition has always proven to be a powerful tool when the identification of an individual comes into play. Therefore the face recognition among newborn is implemented in the proposed system. In order to deploy face recognition system for newborns, first a database is generated maintaining the images of a newborn and all the suspected parents. Matlab is used as a programming tool for the proposed work. The system trains the sample images of the parents using the HMM with a combination of SVD coefficients. The HMM and SVD models provided an approach to model a system that can develop a training sample of the image and can detect the image while any test sample is presented to the program. As a pre-processing method two-dimensional order static filtering is applied to the test images that improve the computational speed and accuracy of the system. Quantization SVD differentiates each image into a sequence of block and then each block of image is regarded as a numeric string. The HMM can effortlessly model the numeric block for training and recognition purpose. Matlab program generates a GUI that can train and recognize the image of a baby’s parent. The proposed system is user friendly and very quick in generating results. A fast and efficient system had been developed for the purpose of face recognition among the newborns

A Quantitative Network Situation Assessment Method Based on Stochastic Model

This paper describes the basic models of network security state evaluation system and concentrates on researching the situation assessment method with stochastic model. In the paper, which makes use of the Hidden Semi-Markov Model (HsMM), tries to simulate the operation of network system. The alert statistics, deriving from network defense system, is used as data sources to realize the evaluation of network security situation. HsMM modifies the HMM model concerning the hypothesis of some state-duration time in relation to exponential distribution, which coincides the description of the network systems operation in the real world. The experimental results imply that HsMM is an ideal security evaluation method.

Research on Associative Memory Models of Emotional Robots

Associative memory is essential to realize man-machine cooperation in the natural interaction between humans and robots. The establishment of associative memory model is to solve the problem. First, based on the theory of emotional energy, mood spontaneous metastasis model and stimulate metastasis model are put forward. Then we can achieve affective computing on the external excitation combining with Markov chain model which is about emotions of spontaneous metastasis and HMM model which is about stimulating metastasis. Second, based on the neural network, the associative memory model which is applied in emotional robots is put forward by calculating the emotional state of the robot's dynamic change of mind and considering their own needs at the same time. Finally, the model was applied to the emotional robot platform which we developed. The effect is validated better.

3D hand gesture tracking and recognition for controlling an intelligent wheelchair

Hand gesture recognition is a user-friendly and intuitive means for human machine interaction. This paper proposes a novel 3D hand gesture recognition method for controlling an intelligent wheelchair based on both colour and depth information. Image depth information of human palm is obtained by a 3D Kinect vision sensor and then its position is obtained through the hand analysis module in OpenNI. The improved Centroid Distance Function is used to extract 3D hand trajectory features, while hidden Markov model HMM is applied to train samples and recognise hand gesture trajectories. Finally, the recognition results are converted into control commands through an ad hoc network and sent to an intelligent wheelchair for its motion control. Experiment results show that the proposed method has good invariance to lighting changes, hand rotation and scaling conditions and is very robust to background interference.

AB INITIO HUMAN miRNA AND PRE-miRNA PREDICTION

MicroRNAs (miRNAs) are small single-stranded noncoding RNAs that play an important role in post-transcriptional regulation of gene expression. In this paper, we present a web server for ab initio prediction of the human miRNAs and their precursors. The prediction methods are based on the hidden Markov Models and the context-structural characteristics. By taking into account the identified patterns of primary and secondary structures of the pre-miRNAs, a new HMM model is proposed and the existing context-structural Markov model is modified. The evaluation of the method performance has shown that it can accurately predict novel human miRNAs. Comparing with the existing methods we demonstrate that our method has a higher prediction quality both for human pre-miRNAs and miRNAs. The models have also showed good results in the prediction of the mouse miRNAs. The web server is available at http://wwwmgs.bionet.nsc.ru/mgs/programs/rnaanalys (mirror http://miRNA.at.nsu.ru ).

Information Extraction from Chinese Papers Based on Hidden Markov Model

Hidden Markov model HMM (1) is one of the important approaches for information extraction. In this paper, a model of the improved first-order hidden Markov HMM (2) is proposed. In the HMM (2), the output probability of the observation is not only dependent on the current state of the model, but also dependent on the previous state of the current state of the model. The algorithm of the ML and the algorithm of the Viterbi are analyzed. At last, experiments show that the HMM (2) is more precise than the HMM (1).

Design and Implementation of Speech Recognition System Based on FPGA

Speech recognition technology is one of the hottest and the most promising new information technologies in the world. This paper studied the voice pretreatment and extractions of MFCC characteristic parameters, constructed speech keywords recognition algorithm with the core of the VQ model and the HMM model, using MATLAB to complete the training and simulation of algorithm, FPGA-based voice recognition technology, and the simulation and implementation of its hardware and software. It laid the foundation for the realization of speech recognition and control based FPGA.

The Phoneme Automatic Segmentation Algorithms Study of Tibetan Lhasa Words Continuous Speech Stream

In this paper, we adopt two methods to voice phoneme segmentation when building Tibetan corpus: One is the traditional artificial segmentation method, one is the automatic segmentation method based on the Mono prime HMM model. And experiments are performed to analyze the accuracy of both methods of segmentations. The results showed: Automatic segmentation method based tone prime HMM model helps to shorten the cycle of building Tibetan corpus, especially in building a large corpus segmentation and labeling a lot of time and manpower cost savings, and have greatly improved the accuracy and consistency of speech corpus annotation information.

The Summarize of Improved HMM Model

The hidden markov model is a kind of important probability model of series data processing and statistical learning and it has been successfully applied in many engineering tasks. This paper introduces the basic principle of hidden markov model firstly, and then discusses the limitations of hidden markov model, as well as the improved hidden markov model which is put forward to solve these problems.

Improvement of Recognition Performance for Limabeam Algorithm by using MLLR Adaptation

This paper presents a method using Maximum-Likelihood Linear Regression (MLLR) adaptation to improve recognition performance of Limabeam algorithm for speech recognition using microphone array. From our investigation on Limabeam algorithm, we can see that the performance of filtering optimization depends strongly on the supporting optimal state sequence and this sequence is created by using Viterbi algorithm trained with HMM model. So we propose an approach using MLLR adaptation for the recognition of speech uttered in a new environment to obtain better optimal state sequence that support for the filtering parameters' optimal step. Experimental results show that the system embedded with MLLR adaptation presents the word correct recognition rate 2% higher than that of original calibrate Limabeam and also present 7% higher than that of Delay and Sum algorithm. The best recognition accuracy of 89.4% is obtained when we use 4 microphones with 5 utterances for adaptation.