The Experts below are selected from a list of 39 Experts worldwide ranked by ideXlab platform
John Yearwood - One of the best experts on this subject based on the ideXlab platform.
-
A Constraint-Based Evolutionary Learning Approach to the Expectation Maximization for Optimal Estimation of the Hidden Markov Model for Speech Signal Modeling
IEEE Transactions on Systems Man and Cybernetics Part B (Cybernetics), 2009Co-Authors: Shamsul Huda, John Yearwood, Roberto TogneriAbstract:This paper attempts to overcome the tendency of the expectation-maximization (EM) algorithm to locate a local rather than global maximum when applied to estimate the hidden Markov model (HMM) parameters in speech signal modeling. We propose a hybrid algorithm for estimation of the HMM in automatic speech recognition (ASR) using a constraint-based evolutionary algorithm (EA) and EM, the CEL-EM. The novelty of our hybrid algorithm (CEL-EM) is that it is applicable for estimation of the constraint-based models with many constraints and large numbers of parameters (which use EM) like HMM. Two constraint-based versions of the CEL-EM with different fusion strategies have been proposed using a constraint-based EA and the EM for better estimation of HMM in ASR. The first one uses a traditional constraint-handling mechanism of EA. The other version transforms a constrained optimization problem into an unconstrained problem using Lagrange multipliers. Fusion strategies for the CEL-EM use a staged-fusion approach where EM has been plugged with the EA periodically after the execution of EA for a specific period of time to maintain the global sampling capabilities of EA in the hybrid algorithm. A Variable Initialization approach (VIA) has been proposed using a Variable segmentation to provide a better Initialization for EA in the CEL-EM. Experimental results on the TIMIT speech corpus show that CEL-EM obtains higher recognition accuracies than the traditional EM algorithm as well as a top-standard EM (VIA-EM, constructed by applying the VIA to EM).
-
a Variable Initialization approach to the em algorithm for better estimation of the parameters of hidden markov model based acoustic modeling of speech signals
International Conference on Data Mining, 2006Co-Authors: Md Shamsul Huda, Ranadhir Ghosh, John YearwoodAbstract:The traditional method for estimation of the parameters of Hidden Markov Model (HMM) based acoustic modeling of speech uses the Expectation-Maximization (EM) algorithm. The EM algorithm is sensitive to initial values of HMM parameters and is likely to terminate at a local maximum of likelihood function resulting in non-optimized estimation for HMM and lower recognition accuracy. In this paper, to obtain better estimation for HMM and higher recognition accuracy, several candidate HMMs are created by applying EM on multiple initial models. The best HMM is chosen from the candidate HMMs which has highest value for likelihood function. Initial models are created by varying maximum frame number in the segmentation step of HMM Initialization process. A binary search is applied while creating the initial models. The proposed method has been tested on TIMIT database. Experimental results show that our approach obtains improved values for likelihood function and improved recognition accuracy.
-
Industrial Conference on Data Mining - A Variable Initialization approach to the EM algorithm for better estimation of the parameters of hidden markov model based acoustic modeling of speech signals
Advances in Data Mining. Applications in Medicine Web Mining Marketing Image and Signal Mining, 2006Co-Authors: Shamsul Huda, Ranadhir Ghosh, John YearwoodAbstract:The traditional method for estimation of the parameters of Hidden Markov Model (HMM) based acoustic modeling of speech uses the Expectation-Maximization (EM) algorithm. The EM algorithm is sensitive to initial values of HMM parameters and is likely to terminate at a local maximum of likelihood function resulting in non-optimized estimation for HMM and lower recognition accuracy. In this paper, to obtain better estimation for HMM and higher recognition accuracy, several candidate HMMs are created by applying EM on multiple initial models. The best HMM is chosen from the candidate HMMs which has highest value for likelihood function. Initial models are created by varying maximum frame number in the segmentation step of HMM Initialization process. A binary search is applied while creating the initial models. The proposed method has been tested on TIMIT database. Experimental results show that our approach obtains improved values for likelihood function and improved recognition accuracy.
Shamsul Huda - One of the best experts on this subject based on the ideXlab platform.
-
A Constraint-Based Evolutionary Learning Approach to the Expectation Maximization for Optimal Estimation of the Hidden Markov Model for Speech Signal Modeling
IEEE Transactions on Systems Man and Cybernetics Part B (Cybernetics), 2009Co-Authors: Shamsul Huda, John Yearwood, Roberto TogneriAbstract:This paper attempts to overcome the tendency of the expectation-maximization (EM) algorithm to locate a local rather than global maximum when applied to estimate the hidden Markov model (HMM) parameters in speech signal modeling. We propose a hybrid algorithm for estimation of the HMM in automatic speech recognition (ASR) using a constraint-based evolutionary algorithm (EA) and EM, the CEL-EM. The novelty of our hybrid algorithm (CEL-EM) is that it is applicable for estimation of the constraint-based models with many constraints and large numbers of parameters (which use EM) like HMM. Two constraint-based versions of the CEL-EM with different fusion strategies have been proposed using a constraint-based EA and the EM for better estimation of HMM in ASR. The first one uses a traditional constraint-handling mechanism of EA. The other version transforms a constrained optimization problem into an unconstrained problem using Lagrange multipliers. Fusion strategies for the CEL-EM use a staged-fusion approach where EM has been plugged with the EA periodically after the execution of EA for a specific period of time to maintain the global sampling capabilities of EA in the hybrid algorithm. A Variable Initialization approach (VIA) has been proposed using a Variable segmentation to provide a better Initialization for EA in the CEL-EM. Experimental results on the TIMIT speech corpus show that CEL-EM obtains higher recognition accuracies than the traditional EM algorithm as well as a top-standard EM (VIA-EM, constructed by applying the VIA to EM).
-
Industrial Conference on Data Mining - A Variable Initialization approach to the EM algorithm for better estimation of the parameters of hidden markov model based acoustic modeling of speech signals
Advances in Data Mining. Applications in Medicine Web Mining Marketing Image and Signal Mining, 2006Co-Authors: Shamsul Huda, Ranadhir Ghosh, John YearwoodAbstract:The traditional method for estimation of the parameters of Hidden Markov Model (HMM) based acoustic modeling of speech uses the Expectation-Maximization (EM) algorithm. The EM algorithm is sensitive to initial values of HMM parameters and is likely to terminate at a local maximum of likelihood function resulting in non-optimized estimation for HMM and lower recognition accuracy. In this paper, to obtain better estimation for HMM and higher recognition accuracy, several candidate HMMs are created by applying EM on multiple initial models. The best HMM is chosen from the candidate HMMs which has highest value for likelihood function. Initial models are created by varying maximum frame number in the segmentation step of HMM Initialization process. A binary search is applied while creating the initial models. The proposed method has been tested on TIMIT database. Experimental results show that our approach obtains improved values for likelihood function and improved recognition accuracy.
Roberto Togneri - One of the best experts on this subject based on the ideXlab platform.
-
A Constraint-Based Evolutionary Learning Approach to the Expectation Maximization for Optimal Estimation of the Hidden Markov Model for Speech Signal Modeling
IEEE Transactions on Systems Man and Cybernetics Part B (Cybernetics), 2009Co-Authors: Shamsul Huda, John Yearwood, Roberto TogneriAbstract:This paper attempts to overcome the tendency of the expectation-maximization (EM) algorithm to locate a local rather than global maximum when applied to estimate the hidden Markov model (HMM) parameters in speech signal modeling. We propose a hybrid algorithm for estimation of the HMM in automatic speech recognition (ASR) using a constraint-based evolutionary algorithm (EA) and EM, the CEL-EM. The novelty of our hybrid algorithm (CEL-EM) is that it is applicable for estimation of the constraint-based models with many constraints and large numbers of parameters (which use EM) like HMM. Two constraint-based versions of the CEL-EM with different fusion strategies have been proposed using a constraint-based EA and the EM for better estimation of HMM in ASR. The first one uses a traditional constraint-handling mechanism of EA. The other version transforms a constrained optimization problem into an unconstrained problem using Lagrange multipliers. Fusion strategies for the CEL-EM use a staged-fusion approach where EM has been plugged with the EA periodically after the execution of EA for a specific period of time to maintain the global sampling capabilities of EA in the hybrid algorithm. A Variable Initialization approach (VIA) has been proposed using a Variable segmentation to provide a better Initialization for EA in the CEL-EM. Experimental results on the TIMIT speech corpus show that CEL-EM obtains higher recognition accuracies than the traditional EM algorithm as well as a top-standard EM (VIA-EM, constructed by applying the VIA to EM).
J. Schumann - One of the best experts on this subject based on the ideXlab platform.
-
Adding assurance to automatically generated code
Eighth IEEE International Symposium on High Assurance Systems Engineering 2004. Proceedings., 2004Co-Authors: E. Denney, B. Fischer, J. SchumannAbstract:Code to estimate position and attitude of a spacecraft or aircraft belongs to the most safety-critical parts of flight software. The complex underlying mathematics and abundance of design details make it error-prone and reliable implementations costly. AutoFilter is a program synthesis tool for the automatic generation of state estimation code from compact specifications. It can automatically produce additional safety certificates which formally guarantee that each generated program individually satisfies a set of important safety policies. These safety policies (e.g., array-bounds, Variable Initialization) form a core of properties which are essential for high-assurance software. Here we describe the AutoFilter system and its certificate generator and compare our approach to the static analysis tool PolySpace.
-
Certification support for automatically generated programs
36th Annual Hawaii International Conference on System Sciences 2003. Proceedings of the, 2003Co-Authors: J. Schumann, B. Fischer, M. Whalen, J. WhittleAbstract:Although autocoding techniques promise large gains in software development productivity, their "real-world" application has been limited, particularly in safety-critical domains. Often, the major impediment is the missing trustworthiness of these systems: demonstrating - let alone formally certifying - the trustworthiness of automatic code generators is extremely difficult due to their complexity and size. We develop an alternative product-oriented certification approach which is based on five principles: (1) trustworthiness of the generator is reduced to the safety of each individual generated program; (2) program safety is defined as adherence to an explicitly formulated safety policy; (3) the safety policy is formalized by a collection of logical program properties; (4) Hoare-style program verification is used to show that each generated program satisfies the required properties; (5) the code generator itself is extended to automatically produce the code annotations required for verification. The approach is feasible because the code generator has full knowledge about the program under construction and about the properties to be verified. It can thus generate all auxiliary code annotations a theorem prover needs to discharge all emerging verification obligations fully automatically. Here we report how this approach is used in a certification extension for AutoBayes, an automatic program synthesis system which generates data analysis programs (e.g., for clustering and time-series analysis) from declarative specifications. In particular, we describe how a Variable-Initialization-before-use safety policy can be encoded and certified.
Hiroki Nomiya - One of the best experts on this subject based on the ideXlab platform.
-
IIAI-AAI - Improvement of the Success Rate of Automatic Generation of Procedural Programs with Variable Initialization Using Genetic Programming
2014 IIAI 3rd International Conference on Advanced Applied Informatics, 2014Co-Authors: Fuuki Horii, Teruhisa Hochin, Hiroki NomiyaAbstract:Genetic Programming (GP), a method of evolutionary computation, is used in producing a variety of programs. In order to generate a procedural program, handling Variables is required. It increases the number of combinations of generated programs. This paper proposes a method including the automatic Initialization of Variables and decreasing the number of combinations of them. For this propose, two major revisions are introduced. One is the introduction of new parameters, the maximum depth and the minimum depth of the height of a program tree. These make programs easy to have a specific structure. The other is the addition of genetic operations. These are for avoiding convergence of programs. Owing to these revisions, it is possible to improve the success rate of the generation of program that includes all of requirement.
-
Improvement of the Success Rate of Automatic Generation of Procedural Programs with Variable Initialization Using Genetic Programming
2014 IIAI 3rd International Conference on Advanced Applied Informatics, 2014Co-Authors: Fuuki Horii, Teruhisa Hochin, Hiroki NomiyaAbstract:Genetic Programming (GP), a method of evolutionary computation, is used in producing a variety of programs. In order to generate a procedural program, handling Variables is required. It increases the number of combinations of generated programs. This paper proposes a method including the automatic Initialization of Variables and decreasing the number of combinations of them. For this propose, two major revisions are introduced. One is the introduction of new parameters, the maximum depth and the minimum depth of the height of a program tree. These make programs easy to have a specific structure. The other is the addition of genetic operations. These are for avoiding convergence of programs. Owing to these revisions, it is possible to improve the success rate of the generation of program that includes all of requirement.
