The Experts below are selected from a list of 324 Experts worldwide ranked by ideXlab platform
Jean-francois Bonastre - One of the best experts on this subject based on the ideXlab platform.
-
Applying SVMs and weight-based Factor Analysis to unsupervised adaptation for speaker verification
Computer Speech and Language, 2011Co-Authors: Mitchell Mclaren, Driss Matrouf, Robbie Vogt, Jean-francois BonastreAbstract:This paper presents an extended study on the implementation of support vector machine (SVM) based speaker verification in systems that employ continuous progressive Model adaptation using the weight-based Factor Analysis Model. The weight-based Factor Analysis Model compensates for session variations in unsupervised scenarios by incorporating trial confidence measures in the general statistics used in the inter-session variability Modelling process. Employing weight-based Factor Analysis in Gaussian mixture Models (GMMs) was recently found to provide significant performance gains to unsupervised classification. Further improvements in performance were found through the integration of SVM-based classification in the system by means of GMM supervectors. This study focuses particularly on the way in which a client is represented in the SVM kernel space using single and multiple target supervectors. Experimental results indicate that training client SVMs using a single target supervector maximises performance while exhibiting a certain robustness to the inclusion of impostor training data in the Model. Furthermore, the inclusion of low-scoring target trials in the adaptation process is investigated where they were found to significantly aid performance.
-
Applying SVMs and weight-based Factor Analysis to unsupervised adaptation for speaker verification
Computer Speech and Language, 2011Co-Authors: Mitchell Mclaren, Driss Matrouf, Robbie Vogt, Jean-francois BonastreAbstract:This paper presents an extended study on the implementation of support vector machine (SVM) based speaker verification in systems that employ continuous progressive Model adaptation using the weight-based Factor Analysis Model. The weight-based Factor Analysis Model compensates for session variations in unsupervised scenarios by incorporating trial confidence measures in the general statistics used in the inter-session variability Modelling process. Employing weight-based Factor Analysis in Gaussian mixture Models (GMMs) was recently found to provide significant performance gains to unsupervised classification. Further improvements in performance were found through the integration of SVM-based classification in the system by means of GMM supervectors. This study focuses particularly on the way in which a client is represented in the SVM kernel space using single and multiple target supervectors. Experimental results indicate that training client SVMs using a single target supervector maximises performance while exhibiting a certain robustness to the inclusion of impostor training data in the Model. Furthermore, the inclusion of low-scoring target trials in the adaptation process is investigated where they were found to significantly aid performance.
-
a straightforward and efficient implementation of the Factor Analysis Model for speaker verification
Conference of the International Speech Communication Association, 2007Co-Authors: Driss Matrouf, Nicolas Scheffer, Benoît Fauve, Jean-francois BonastreAbstract:For a few years, the problem of session variability in textindependent automatic speaker verification is being tackled actively. A new paradigm based on a Factor Analysis Model have successfully been applied for this task. While very efficient, its implementation is demanding. In this paper, the algorithms involved in the eigenchannel MAP Model are written down for a straightforward implementation, without referring to previous work or complex mathematics. In addition, a different compensation scheme is proposed where the standard GMM likelihood can be used without any modification to obtain good performance (even without the need of score normalization). The use of the compensated supervectors within a SVM classifier through a distance based kernel is also investigated. Experiments results shows an overall 50% relative gain over the standard GMM-UBM system on NIST SRE 2005 and 2006 protocols (both at the DCFmin and EER). Index Terms: Speaker Verification, Session variability, Factor Analysis, GMM Supervectors.
-
A Straightforward and Efficient Implementation of the Factor Analysis Model for Speaker Verification
2007Co-Authors: Driss Matrouf, Nicolas Scheffer, Benoît Fauve, Jean-francois BonastreAbstract:For a few years, the problem of session variability in text-independent automatic speaker verification is being tackled actively. A new paradigm based on a Factor Analysis Model have successfully been applied for this task. While very efficient, its implementation is demanding. In this paper, the algorithms involved in the eigenchannel MAP Model are written down for a straightforward implementation, without referring to previous work or complex mathematics. In addition, a different compensation scheme is proposed where the standard GMM likelihood can be used without any modification to obtain good performance (even without the need of score normalization). The use of the compensated supervectors within a SVM classifier through a distance based kernel is also investigated. Experiments results shows an overall 50% relative gain over the standard GMM-UBM system on NIST SRE 2005 and 2006 protocols (both at the DCFmin and EER).
-
INTERSPEECH - A Straightforward and Efficient Implementation of the Factor Analysis Model for Speaker Verification
2007Co-Authors: Driss Matrouf, Nicolas Scheffer, Benoît Fauve, Jean-francois BonastreAbstract:For a few years, the problem of session variability in textindependent automatic speaker verification is being tackled actively. A new paradigm based on a Factor Analysis Model have successfully been applied for this task. While very efficient, its implementation is demanding. In this paper, the algorithms involved in the eigenchannel MAP Model are written down for a straightforward implementation, without referring to previous work or complex mathematics. In addition, a different compensation scheme is proposed where the standard GMM likelihood can be used without any modification to obtain good performance (even without the need of score normalization). The use of the compensated supervectors within a SVM classifier through a distance based kernel is also investigated. Experiments results shows an overall 50% relative gain over the standard GMM-UBM system on NIST SRE 2005 and 2006 protocols (both at the DCFmin and EER). Index Terms: Speaker Verification, Session variability, Factor Analysis, GMM Supervectors.
Yu-ting Chen - One of the best experts on this subject based on the ideXlab platform.
-
An effective taxi recommender system based on a spatio-temporal Factor Analysis Model
Information Sciences, 2015Co-Authors: Ren-hung Hwang, Yu-ling Hsueh, Yu-ting ChenAbstract:We investigate four Factors for recommending a potential high-revenue location.The OFF-ON Model obtains the relation between the drop-off and the get-on locations.The ON-OFF Model estimates the fare for a trip from the recommended location.Results show the revenue on weekdays is better than that on weekends.Distance to the next cruising location and waiting time are important features. The taxi fleet management systems based on GPS have become an important tool for taxi businesses. Such systems can be used not only for fleet management, but also to provide useful information for taxi drivers to increase their profits by mining historical GPS trajectories. In this paper, we propose a taxi recommender system for determining the next cruising location, which could be a value-added module in fleet management systems. In the literature, three Factors have been considered in different studies to address a similar objective: distance between the current location and the recommended location, waiting time for the next passengers, and expected fare for the trip. In this paper, in addition to these Factors, we consider one key Factor based on driver experience: what is the most likely location to pick up passengers, given the current passenger drop off location. A location-to-location graph Model, referred to as an OFF-ON Model, is adopted to capture the relation between the passenger drop-off location and the next passenger get-on location. We also adopt an ON-OFF Model to estimate the expected fare for a trip that begins at a recommended location. A real-world dataset from CRAWDAD is used to evaluate the proposed system. A simulator that simulates the cruising behavior of taxies in the dataset and a virtual taxi that cruises based on our recommender system is developed. Our simulation results indicate that although the statistics of the historical data may be different from real-time passenger requests, our recommender system is still effective in terms of recommending more profitable cruising locations.
-
ICNC - An effective taxi recommender system based on a spatiotemporal Factor Analysis Model
2014 International Conference on Computing Networking and Communications (ICNC), 2014Co-Authors: Yu-ling Hsueh, Ren-hung Hwang, Yu-ting ChenAbstract:The taxi fleet management system based on GPS has become an important tool for efficient taxi business. It can be used not only for the sake of fleet management, but also to provide useful information for taxi drivers to earn more profit by mining the historical GPS trajectories. In this paper, we propose a taxi recommender system for next cruising location which could be a value added module of the fleet management system. In the literature, three Factors have been considered in different works to provide the similar objective, which are distance between the current location and the recommended location, waiting time for next passengers, and expected fare for the trip. In this paper, in addition to these Factors, we consider one more Factor based on drivers experience which is the most likely location to pick up passengers given the current passenger drop off location. A location-to-location graph Model, referred to as OFF-ON Model, is adopted to capture the relation between the passenger get-off location and the next passenger get-on location. We also adopted a ON-OFF Model to estimate the expected fare for a trip started at a recommended location. A real world dataset from CRAWDAD is used to evaluated the proposed system. A simulator is developed which simulates cruising behavior of taxies in the dataset and one virtual taxi which cruises based on our recommender system. Our simulation results indicate that although the statistics of historical data may be different from real-time passenger requests, our proposed recommender system is still effective on recommending better profitable cruising location.
Driss Matrouf - One of the best experts on this subject based on the ideXlab platform.
-
Applying SVMs and weight-based Factor Analysis to unsupervised adaptation for speaker verification
Computer Speech and Language, 2011Co-Authors: Mitchell Mclaren, Driss Matrouf, Robbie Vogt, Jean-francois BonastreAbstract:This paper presents an extended study on the implementation of support vector machine (SVM) based speaker verification in systems that employ continuous progressive Model adaptation using the weight-based Factor Analysis Model. The weight-based Factor Analysis Model compensates for session variations in unsupervised scenarios by incorporating trial confidence measures in the general statistics used in the inter-session variability Modelling process. Employing weight-based Factor Analysis in Gaussian mixture Models (GMMs) was recently found to provide significant performance gains to unsupervised classification. Further improvements in performance were found through the integration of SVM-based classification in the system by means of GMM supervectors. This study focuses particularly on the way in which a client is represented in the SVM kernel space using single and multiple target supervectors. Experimental results indicate that training client SVMs using a single target supervector maximises performance while exhibiting a certain robustness to the inclusion of impostor training data in the Model. Furthermore, the inclusion of low-scoring target trials in the adaptation process is investigated where they were found to significantly aid performance.
-
Applying SVMs and weight-based Factor Analysis to unsupervised adaptation for speaker verification
Computer Speech and Language, 2011Co-Authors: Mitchell Mclaren, Driss Matrouf, Robbie Vogt, Jean-francois BonastreAbstract:This paper presents an extended study on the implementation of support vector machine (SVM) based speaker verification in systems that employ continuous progressive Model adaptation using the weight-based Factor Analysis Model. The weight-based Factor Analysis Model compensates for session variations in unsupervised scenarios by incorporating trial confidence measures in the general statistics used in the inter-session variability Modelling process. Employing weight-based Factor Analysis in Gaussian mixture Models (GMMs) was recently found to provide significant performance gains to unsupervised classification. Further improvements in performance were found through the integration of SVM-based classification in the system by means of GMM supervectors. This study focuses particularly on the way in which a client is represented in the SVM kernel space using single and multiple target supervectors. Experimental results indicate that training client SVMs using a single target supervector maximises performance while exhibiting a certain robustness to the inclusion of impostor training data in the Model. Furthermore, the inclusion of low-scoring target trials in the adaptation process is investigated where they were found to significantly aid performance.
-
a straightforward and efficient implementation of the Factor Analysis Model for speaker verification
Conference of the International Speech Communication Association, 2007Co-Authors: Driss Matrouf, Nicolas Scheffer, Benoît Fauve, Jean-francois BonastreAbstract:For a few years, the problem of session variability in textindependent automatic speaker verification is being tackled actively. A new paradigm based on a Factor Analysis Model have successfully been applied for this task. While very efficient, its implementation is demanding. In this paper, the algorithms involved in the eigenchannel MAP Model are written down for a straightforward implementation, without referring to previous work or complex mathematics. In addition, a different compensation scheme is proposed where the standard GMM likelihood can be used without any modification to obtain good performance (even without the need of score normalization). The use of the compensated supervectors within a SVM classifier through a distance based kernel is also investigated. Experiments results shows an overall 50% relative gain over the standard GMM-UBM system on NIST SRE 2005 and 2006 protocols (both at the DCFmin and EER). Index Terms: Speaker Verification, Session variability, Factor Analysis, GMM Supervectors.
-
A Straightforward and Efficient Implementation of the Factor Analysis Model for Speaker Verification
2007Co-Authors: Driss Matrouf, Nicolas Scheffer, Benoît Fauve, Jean-francois BonastreAbstract:For a few years, the problem of session variability in text-independent automatic speaker verification is being tackled actively. A new paradigm based on a Factor Analysis Model have successfully been applied for this task. While very efficient, its implementation is demanding. In this paper, the algorithms involved in the eigenchannel MAP Model are written down for a straightforward implementation, without referring to previous work or complex mathematics. In addition, a different compensation scheme is proposed where the standard GMM likelihood can be used without any modification to obtain good performance (even without the need of score normalization). The use of the compensated supervectors within a SVM classifier through a distance based kernel is also investigated. Experiments results shows an overall 50% relative gain over the standard GMM-UBM system on NIST SRE 2005 and 2006 protocols (both at the DCFmin and EER).
-
INTERSPEECH - A Straightforward and Efficient Implementation of the Factor Analysis Model for Speaker Verification
2007Co-Authors: Driss Matrouf, Nicolas Scheffer, Benoît Fauve, Jean-francois BonastreAbstract:For a few years, the problem of session variability in textindependent automatic speaker verification is being tackled actively. A new paradigm based on a Factor Analysis Model have successfully been applied for this task. While very efficient, its implementation is demanding. In this paper, the algorithms involved in the eigenchannel MAP Model are written down for a straightforward implementation, without referring to previous work or complex mathematics. In addition, a different compensation scheme is proposed where the standard GMM likelihood can be used without any modification to obtain good performance (even without the need of score normalization). The use of the compensated supervectors within a SVM classifier through a distance based kernel is also investigated. Experiments results shows an overall 50% relative gain over the standard GMM-UBM system on NIST SRE 2005 and 2006 protocols (both at the DCFmin and EER). Index Terms: Speaker Verification, Session variability, Factor Analysis, GMM Supervectors.
Ren-hung Hwang - One of the best experts on this subject based on the ideXlab platform.
-
An effective taxi recommender system based on a spatio-temporal Factor Analysis Model
Information Sciences, 2015Co-Authors: Ren-hung Hwang, Yu-ling Hsueh, Yu-ting ChenAbstract:We investigate four Factors for recommending a potential high-revenue location.The OFF-ON Model obtains the relation between the drop-off and the get-on locations.The ON-OFF Model estimates the fare for a trip from the recommended location.Results show the revenue on weekdays is better than that on weekends.Distance to the next cruising location and waiting time are important features. The taxi fleet management systems based on GPS have become an important tool for taxi businesses. Such systems can be used not only for fleet management, but also to provide useful information for taxi drivers to increase their profits by mining historical GPS trajectories. In this paper, we propose a taxi recommender system for determining the next cruising location, which could be a value-added module in fleet management systems. In the literature, three Factors have been considered in different studies to address a similar objective: distance between the current location and the recommended location, waiting time for the next passengers, and expected fare for the trip. In this paper, in addition to these Factors, we consider one key Factor based on driver experience: what is the most likely location to pick up passengers, given the current passenger drop off location. A location-to-location graph Model, referred to as an OFF-ON Model, is adopted to capture the relation between the passenger drop-off location and the next passenger get-on location. We also adopt an ON-OFF Model to estimate the expected fare for a trip that begins at a recommended location. A real-world dataset from CRAWDAD is used to evaluate the proposed system. A simulator that simulates the cruising behavior of taxies in the dataset and a virtual taxi that cruises based on our recommender system is developed. Our simulation results indicate that although the statistics of the historical data may be different from real-time passenger requests, our recommender system is still effective in terms of recommending more profitable cruising locations.
-
ICNC - An effective taxi recommender system based on a spatiotemporal Factor Analysis Model
2014 International Conference on Computing Networking and Communications (ICNC), 2014Co-Authors: Yu-ling Hsueh, Ren-hung Hwang, Yu-ting ChenAbstract:The taxi fleet management system based on GPS has become an important tool for efficient taxi business. It can be used not only for the sake of fleet management, but also to provide useful information for taxi drivers to earn more profit by mining the historical GPS trajectories. In this paper, we propose a taxi recommender system for next cruising location which could be a value added module of the fleet management system. In the literature, three Factors have been considered in different works to provide the similar objective, which are distance between the current location and the recommended location, waiting time for next passengers, and expected fare for the trip. In this paper, in addition to these Factors, we consider one more Factor based on drivers experience which is the most likely location to pick up passengers given the current passenger drop off location. A location-to-location graph Model, referred to as OFF-ON Model, is adopted to capture the relation between the passenger get-off location and the next passenger get-on location. We also adopted a ON-OFF Model to estimate the expected fare for a trip started at a recommended location. A real world dataset from CRAWDAD is used to evaluated the proposed system. A simulator is developed which simulates cruising behavior of taxies in the dataset and one virtual taxi which cruises based on our recommender system. Our simulation results indicate that although the statistics of historical data may be different from real-time passenger requests, our proposed recommender system is still effective on recommending better profitable cruising location.
Yu-ling Hsueh - One of the best experts on this subject based on the ideXlab platform.
-
An effective taxi recommender system based on a spatio-temporal Factor Analysis Model
Information Sciences, 2015Co-Authors: Ren-hung Hwang, Yu-ling Hsueh, Yu-ting ChenAbstract:We investigate four Factors for recommending a potential high-revenue location.The OFF-ON Model obtains the relation between the drop-off and the get-on locations.The ON-OFF Model estimates the fare for a trip from the recommended location.Results show the revenue on weekdays is better than that on weekends.Distance to the next cruising location and waiting time are important features. The taxi fleet management systems based on GPS have become an important tool for taxi businesses. Such systems can be used not only for fleet management, but also to provide useful information for taxi drivers to increase their profits by mining historical GPS trajectories. In this paper, we propose a taxi recommender system for determining the next cruising location, which could be a value-added module in fleet management systems. In the literature, three Factors have been considered in different studies to address a similar objective: distance between the current location and the recommended location, waiting time for the next passengers, and expected fare for the trip. In this paper, in addition to these Factors, we consider one key Factor based on driver experience: what is the most likely location to pick up passengers, given the current passenger drop off location. A location-to-location graph Model, referred to as an OFF-ON Model, is adopted to capture the relation between the passenger drop-off location and the next passenger get-on location. We also adopt an ON-OFF Model to estimate the expected fare for a trip that begins at a recommended location. A real-world dataset from CRAWDAD is used to evaluate the proposed system. A simulator that simulates the cruising behavior of taxies in the dataset and a virtual taxi that cruises based on our recommender system is developed. Our simulation results indicate that although the statistics of the historical data may be different from real-time passenger requests, our recommender system is still effective in terms of recommending more profitable cruising locations.
-
ICNC - An effective taxi recommender system based on a spatiotemporal Factor Analysis Model
2014 International Conference on Computing Networking and Communications (ICNC), 2014Co-Authors: Yu-ling Hsueh, Ren-hung Hwang, Yu-ting ChenAbstract:The taxi fleet management system based on GPS has become an important tool for efficient taxi business. It can be used not only for the sake of fleet management, but also to provide useful information for taxi drivers to earn more profit by mining the historical GPS trajectories. In this paper, we propose a taxi recommender system for next cruising location which could be a value added module of the fleet management system. In the literature, three Factors have been considered in different works to provide the similar objective, which are distance between the current location and the recommended location, waiting time for next passengers, and expected fare for the trip. In this paper, in addition to these Factors, we consider one more Factor based on drivers experience which is the most likely location to pick up passengers given the current passenger drop off location. A location-to-location graph Model, referred to as OFF-ON Model, is adopted to capture the relation between the passenger get-off location and the next passenger get-on location. We also adopted a ON-OFF Model to estimate the expected fare for a trip started at a recommended location. A real world dataset from CRAWDAD is used to evaluated the proposed system. A simulator is developed which simulates cruising behavior of taxies in the dataset and one virtual taxi which cruises based on our recommender system. Our simulation results indicate that although the statistics of historical data may be different from real-time passenger requests, our proposed recommender system is still effective on recommending better profitable cruising location.
