Recent proliferation of sensor networks in various application areas has promoted real-time behavioral monitoring of various physical systems. The opportunity to use sensor generated data dynamically for improving spe...
详细信息
Recent proliferation of sensor networks in various application areas has promoted real-time behavioral monitoring of various physical systems. The opportunity to use sensor generated data dynamically for improving speed, accuracy, and general performance of predictive behavior modeling simulation is of paramount importance. The present paper identifies enabling modeling methods and computational strategies that are critical for achieving real-time or near real-time simulation response of very large and complex systems. It also discusses our choices of these technologies in the context of sample multidisciplinary computational mechanics applications and describes two examples to demonstrate the feasibility of integrating real-time data with real-time simulation.
To reduce speech recognition error rate we can use better statistical language models. These models can be improved by grouping words into word equivalence classes. Clustering algorithms can be used to automatically d...
详细信息
To reduce speech recognition error rate we can use better statistical language models. These models can be improved by grouping words into word equivalence classes. Clustering algorithms can be used to automatically do this word grouping. We present an incremental clustering algorithm and two iterative clustering algorithms. Also, we compare them with previous algorithms. The experimental results show that the two iterative algorithms perform as well as previous ones. It should be pointed out that one of them, that uses the leaving one out technique, has the ability to automatically determine the optimum number of classes. These iterative algorithms are used by the incremental one. On the other hand, the proposed incremental algorithm achieves the best results of the compared algorithms, its behavior is the most regular with the variation of the number of classes and can automatically determine the optimum number of classes.
This paper studies speech-driven Web retrieval models which accepts spoken search topics (queries) in the NTCIR-3 Web retrieval task. The major focus of this paper is on improving speech recognition accuracy of spoken...
详细信息
This paper studies speech-driven Web retrieval models which accepts spoken search topics (queries) in the NTCIR-3 Web retrieval task. The major focus of this paper is on improving speech recognition accuracy of spoken queries then improving retrieval accuracy in speech-driven Web retrieval. We experimentally evaluate the techniques of combining outputs of multiple LVCSR models in recognition of spoken queries. As model combination techniques, we compare the SVM learning technique conventional voting schemes such as ROVER. We show that the techniques of multiple LVCSR model combination can achieve improvement both in speech recognition and retrieval accuracies in speech-driven text retrieval. We also show that model combination by SVM learning outperforms conventional voting schemes both in speech recognition retrieval accuracies.
The growing volume of heterogeneous and distributed educational resources on the WWW makes difficult for the existing tools to retrieve relevant information. To improve the performance of these tools, we describe a fl...
详细信息
ISBN:
(纸本)088986361X
The growing volume of heterogeneous and distributed educational resources on the WWW makes difficult for the existing tools to retrieve relevant information. To improve the performance of these tools, we describe a flexible architecture based on two kinds of robots : "generalists" and "specialists" that collect and organize these metadata, in order to localize the resources on the WWW. They will contribute to the overall autoorganizing information process by exchanging their indices.
Into the framework of the Almanzora*GIS project, which objective was building a GIS tool to develop two counties north the province of Almeria, it was required to build an insolation model to evaluate proposals to sto...
详细信息
Into the framework of the Almanzora*GIS project, which objective was building a GIS tool to develop two counties north the province of Almeria, it was required to build an insolation model to evaluate proposals to stop the erosion of the area. Due to the lack of detailed field data and the complex topography of the area, we decided to compute the insolation theoretically using the SolarFlux macro with ARC/Info. Computing insolation limits along the year was relatively easy using the solstices values, but evaluating the received insolation along the year, taking account of the relief required a long computing time. In this work, we present how we evaluated the impact, in quality of data and computing time, of taking account of the relief in the study area and their environments, as well as the distribution of the influence of the relief. This study presents a methodology and an example of the analysis of the conditions to compute the insolation maps taking account of the relief conditions.
In the majority of analytical and imitation models analyzing the processes of communication networks functioning, input streams are associated with the well known Poisson's model, describing the most unfavorable c...
详细信息
In the majority of analytical and imitation models analyzing the processes of communication networks functioning, input streams are associated with the well known Poisson's model, describing the most unfavorable case of stationary random streams. In practice the input streams are characterized mainly with strong nonstationary processes, impacting the final results of modeling to a significant degree. Having this in mind this report presents an input stream model for simulation modeling and analysis of communication networks.
An overview is given of a formal framework and specific implementations which address the problems that arise when an untrusted third party publisher collects and organizes data from many different data owners and the...
详细信息
An overview is given of a formal framework and specific implementations which address the problems that arise when an untrusted third party publisher collects and organizes data from many different data owners and then provides answers to user queries on the combined data set. With this scheme, owners can be confident their data is properly represented and users can be confident they have correct answers. It is shown that a group of data owners can efficiently certify that an untrusted third party publisher has computed the correct digest of the owners' collected data sets.
We study mathematical models and discuss optimization algorithms for the dimensioning of 3G multimedia networks. We propose two models which aims at dimensioning networks with both a radio and a core component. The fi...
详细信息
We study mathematical models and discuss optimization algorithms for the dimensioning of 3G multimedia networks. We propose two models which aims at dimensioning networks with both a radio and a core component. The first one is an anticipative one in which we assume that we know a priori the traffic over the planning period and the dimensioning is defined with a best possible call admission control procedure. The second one is a causal one in which we define an explicit call admission control procedure which makes the accept/reject decisions without any knowledge on the forthcoming traffic. We then compare, on an experimental basis, the dimensioning obtained by both models on some multi-service networks.
For many practical applications of speech recognition systems, it is quite desirable to have an estimate of confidence for each hypothesized word. Unlike previous works on confidence measures, we have proposed feature...
详细信息
ISBN:
(纸本)0780376633
For many practical applications of speech recognition systems, it is quite desirable to have an estimate of confidence for each hypothesized word. Unlike previous works on confidence measures, we have proposed features for confidence measures that are extracted from outputs of more than one LVCSR models. For further analysis of the proposed confidence measure, this paper examines the correlation between each word's confidence and the word's features such as its part-of-speech and syllable length. We then apply SVM learning technique to the task of combining outputs of multiple LVCSR models, where, as features of SVM learning, information such as the pairs of the models which output the hypothesized word are useful for improving the word recognition rate. Experimental results show that the combination results achieve a relative word error reduction of up to 72 % against the best performing single model and that of up to 36 % against ROVER.
Deformable models have been intensively studied in image analysis through the last decade, and are used for detection and recognition of flexible or rigid templates under diverse viewing conditions. Genetic algorithm ...
详细信息
Deformable models have been intensively studied in image analysis through the last decade, and are used for detection and recognition of flexible or rigid templates under diverse viewing conditions. Genetic algorithm (GA) based deformable models are used for generic visual landmark detection and interpretation. The developed system allows topologic localization and navigation using natural and artificial landmarks, exploiting deformable models' ability for handling landmark perspective variations. The resulting perception module has been integrated successfully in a complex navigation system. Various experimental results in real environments are presented on this paper, showing the effectiveness and capacity of the landmark detection and reading system.
暂无评论