Results 1  10
of
13
A tutorial on hidden markov models and selected applications in speech recognition
 Proceedings of the IEEE
, 1989
"... Although initially introduced and studied in the late 1960s and early 1970s, statistical methods of Markov source or hidden Markov modeling have become increasingly popular in the last several years. There are two strong reasons why this has occurred. First the models are very rich in mathematical s ..."
Abstract

Cited by 4251 (1 self)
 Add to MetaCart
Although initially introduced and studied in the late 1960s and early 1970s, statistical methods of Markov source or hidden Markov modeling have become increasingly popular in the last several years. There are two strong reasons why this has occurred. First the models are very rich in mathematical structure and hence can form the theoretical basis for use in a wide range of applications. Second the models, when applied properly, work very well in practice for several important applications. In this paper we attempt to carefully and methodically review the theoretical aspects of this type of statistical modeling and show how they have been applied to selected problems in machine recognition of speech. I.
HMMs and Coupled HMMs for Multichannel EEG Classification
, 2002
"... A variety of Coupled HMMs (CHMMs) have recently been proposed as extensions of HMM to better characterize multiple interdependent sequences. This paper introduces a novel distance coupled HMM. It then compares the performance of several HMM and CHMM models for a multichannel EEG classification prob ..."
Abstract

Cited by 31 (6 self)
 Add to MetaCart
A variety of Coupled HMMs (CHMMs) have recently been proposed as extensions of HMM to better characterize multiple interdependent sequences. This paper introduces a novel distance coupled HMM. It then compares the performance of several HMM and CHMM models for a multichannel EEG classification problem. The results show that, of all approaches examined, the multivariate HMM that has low computational complexity surprisingly outperforms all other models.
Hidden Markov Models as a Process Monitor in Robotic Assembly
, 1996
"... A process monitor for robotic assembly based on Hidden Markov Models (HMMs) is presented. The HMM process monitor is based on the dynamic force/torque signals arising from interaction between the workpiece and the environment. The HMMs represent a stochastic, knowledgebased system where the models ..."
Abstract

Cited by 24 (4 self)
 Add to MetaCart
A process monitor for robotic assembly based on Hidden Markov Models (HMMs) is presented. The HMM process monitor is based on the dynamic force/torque signals arising from interaction between the workpiece and the environment. The HMMs represent a stochastic, knowledgebased system where the models are trained offline with the BaumWelch reestimation algorithm. The assembly task is modeled as a discrete event dynamic system, where a discrete event is defined as a change in contact state between the workpiece and the environment. Our method 1) allows for dynamic motions of the workpiece, 2) accounts for sensor noise and friction and 3) exploits the fact that the amount of force information is large when there is a sudden change of discrete state in robotic assembly. After the HMMs have been trained, we use them online in a 2D experimental setup to recognise discrete events as they occur. Successful event recognition with an accuracy as high as 97% was achieved in 0.50.6 seconds with...
Stochastic Similarity for Validating Human Control Strategy Models
 IEEE Transactions on Robotics and Automation
, 1998
"... Modeling dynamic human control strategy (HCS), or human skill in response to realtime sensing is becoming an increasingly popular paradigm in many different research areas, such as intelligent vehicle systems, virtual reality, and space robotics. Such models are often learned from experimental data ..."
Abstract

Cited by 18 (6 self)
 Add to MetaCart
Modeling dynamic human control strategy (HCS), or human skill in response to realtime sensing is becoming an increasingly popular paradigm in many different research areas, such as intelligent vehicle systems, virtual reality, and space robotics. Such models are often learned from experimental data, and as such can be characterized despite the lack of a good physical model. Unfortunately, learned models presently offer few, if any, guarantees in terms of model fidelity to the training data. This is especially true for dynamic reaction skills, where errors can feed back on themselves to generate state and command trajectories uncharacteristic of the source process. Thus, we propose a stochastic similarity measurebased on hidden Markov model (HMM) analysiscapable of comparing and contrasting stochastic, dynamic, multidimensional trajectories. This similarity measure is the first step in validating a learned model's fidelity to its training data by comparing the model's dynamic trajectories in the feedback loop to the human's dynamic trajectories. In this paper, we first derive and demonstrate properties of the similarity measure for stochastic systems. We then apply the similarity measure to realtime human driving data by comparing different control strategies among different individuals. We show that the proposed similarity measure out performs the more traditional Bayes classifier in correctly grouping driving data from the same individual. Finally, we illustrate how the similarity measure can be used in the validation of models which are learned from experimental data, and how we can connect model validation and model learning to iteratively improve our models of human control strategy.
FrequencyDomain Force Measurements for Discrete Event Contact Recognition
, 1996
"... Discrete event recognition based on force measurements in the frequencydomain is presented. The force signals arise from interaction between the workpiece and the environment in a planar assembly task. The discrete events are modeled as Hidden Markov Models (HMMs), where the models are trained off ..."
Abstract

Cited by 10 (4 self)
 Add to MetaCart
Discrete event recognition based on force measurements in the frequencydomain is presented. The force signals arise from interaction between the workpiece and the environment in a planar assembly task. The discrete events are modeled as Hidden Markov Models (HMMs), where the models are trained offline with the BaumWelch reestimation algorithm. After the HMMs have been trained, we use them online in a robotic system to recognise discrete events as they occur. Event recognition with an accuracy as high as 98% was accomplished in 0.50.6s with a relatively small training set. 1 Introduction Process plants must deal with changing states, multiple faults, unexpected situations and unreliable measurements. To handle these problems realtime process monitoring is essential. Process monitoring is widely used as a component in many industrial processes. In applications such as robotic assembly, however, there is an increasing need for efficient process monitoring methods to account for u...
Semisupervised sequence classification with HMMs
 International Journal of Pattern Recognition & Artificial Intelligence
, 2005
"... Using unlabeled data to help supervised learning has become an increasingly attractive methodology and proven to be effective in many applications. This paper applies semisupervised classification algorithms, based on hidden Markov models (HMMs), to classify sequences. For modelbased classificatio ..."
Abstract

Cited by 5 (1 self)
 Add to MetaCart
Using unlabeled data to help supervised learning has become an increasingly attractive methodology and proven to be effective in many applications. This paper applies semisupervised classification algorithms, based on hidden Markov models (HMMs), to classify sequences. For modelbased classification, semisupervised learning amounts to using both labeled and unlabeled data to train model parameters. We examine three different strategies of using labeled and unlabeled data in the model training process. These strategies differ in how and when labeled and unlabeled data contribute in the whole model training process. Our experimental results on synthetic and real EEG timeseries show that substantially improved classification accuracy can be achieved by these semisupervised learning strategies.
Progress Report: MultiAperture SAR Target Detection Using Hidden Markov Models
, 1994
"... This report highlights our current work and accomplishments on the project to exploit angular diversity for improved target detection in multiaperture SAR images. This report also contains a brief introduction to hidden Markov models, and identifies issues which we will resolve as work continues. W ..."
Abstract

Cited by 1 (1 self)
 Add to MetaCart
This report highlights our current work and accomplishments on the project to exploit angular diversity for improved target detection in multiaperture SAR images. This report also contains a brief introduction to hidden Markov models, and identifies issues which we will resolve as work continues. We have analyzed multiaperture SAR images and demonstrated that anisotropic behavior is present in our multiaperture SAR image set. We performed baselining studies using the common method of CFAR LTT detection and formulated an HMM detection method using BaumWelch reestimation to train HMMs to represent target, tree clutter, and ground clutter pixels. Our results show that HMM detection produced significantly better results than CFAR LTT detection (with a 29by29 reference window) for the y This research was supported by Wright Laboratory. z The SPANN Lab's WWW URL is http://eewww.eng.ohiostate.edu/research/spann/. same multiaperture SAR image and requires less computation. Speci...
Simulated Anealing Approach for Training Hidden Markov Models
"... A simulated annealing method for estimating the parameters of Hidden Markov Models is presented. This method is based on the choice of the optimal trajectory of the discrete state. It is applied to both discrete and continuous observations. The program developped needs no specific initialization of ..."
Abstract
 Add to MetaCart
A simulated annealing method for estimating the parameters of Hidden Markov Models is presented. This method is based on the choice of the optimal trajectory of the discrete state. It is applied to both discrete and continuous observations. The program developped needs no specific initialization of the algorithm by the user, the cooling schedule being general and applicable to any specific model. The method is applied to data generated randomly and compared to the initial model. Numerical experience in using the method is also presented. 1. INTRODUCTION Hidden Markov Models (HMM) have been widely applied in automatic speech recognition. In this field signals are encoded as temporal variation of short time power spectrum [15]. HMM applications are now being extended to many fields such as pattern recognition, signal processing and control. They are well suited for the classification of one or two dimensional signals. An HMM is a double stochastic process with one underlying process tha...
State Transition Recognition in Robotic Assembly Using Hidden Markov Models
, 1995
"... A process monitor for robotic assembly based on Hidden Markov Models (HMMs) is presented. The measurements are the force/torque signals arising from interaction between the workpiece and the environment for a planar assembly task. The HMMs represent a stochastic knowledgebased system where the mode ..."
Abstract
 Add to MetaCart
A process monitor for robotic assembly based on Hidden Markov Models (HMMs) is presented. The measurements are the force/torque signals arising from interaction between the workpiece and the environment for a planar assembly task. The HMMs represent a stochastic knowledgebased system where the models are trained offline with the BaumWelch reestimation algorithm. After the HMMs have been trained, we use them online in a robotic system to recognise events as they occur. Process monitoring with an accuracy of 98% was accomplished in 0.50.6s. 1 Introduction Process plants must deal with changing states, multiple faults, unexpected situations and unreliable measurements. To handle these problems realtime process monitoring is essential. Process monitoring is widely used as a component in many industrial processes. In robotic assembly, however, there is an increasing need for efficient process monitoring methods to account for existing uncertainties of workpieces and the environment...
unknown title
, 2003
"... positive semidefinite matrix V. The figure also shows the set M (z) for three different choices of the vector z 2. in which case the complement of M(z) in Q K denoted by M c (z) is a convex set. Finally, we point out that Theorem 2 does not completely solve the convexity problem in systems with domi ..."
Abstract
 Add to MetaCart
positive semidefinite matrix V. The figure also shows the set M (z) for three different choices of the vector z 2. in which case the complement of M(z) in Q K denoted by M c (z) is a convex set. Finally, we point out that Theorem 2 does not completely solve the convexity problem in systems with dominant selfinterference since the symmetry condition on V is not necessarily satisfied in wireless networks. For the case that V is not symmetric but its diagonal elements are dominant in the sense that Vs =(V+ V T)=2 is positive semidefinite, it seems that F is convex as well if is a convex function. However, we have no proof for that to be true in general. We point out that Vs is the orthogonal projection of V onto the space of K 2 K symmetric matrices. Thus, in this sense, Vs is the closest symmetric matrix to V. This suggests that F is a “nearly convex set ” when both Vs is positive semidefinite (selfinterference dominant) and the distance between V and Vs is sufficiently small. V. CONCLUSION This correspondence addresses the problem of convexity of the feasible QoS region, which is determined by the Perron root of some QoSdependent nonnegative matrix. In particular, the feasible QoS region is convex if the Perron root is a convex function of the QoS vector. By the results of Section III and the previous work, we know that the Perron root is convex, regardless of the number of users and the gain matrix, if and only if each SIR is a logconvex function of the corresponding QoS value. In contrast, when the gain matrix is confined to be symmetric positive semidefinite, we have shown that the Perron root is convex (and with it the feasible QoS region), provided that the SIR is a convex function of the QoS value.