The Experts below are selected from a list of 8865 Experts worldwide ranked by ideXlab platform
Najim Dehak - One of the best experts on this subject based on the ideXlab platform.
-
Odyssey - Cosine Similarity Scoring without Score Normalization Techniques.
2020Co-Authors: Najim Dehak, Reda Dehak, James Glass, Douglas A Reynolds, Patrick KennyAbstract:In recent work [1], a simplified and highly effective approach to speaker recognition based on the Cosine Similarity between lowdimensional vectors, termed ivectors, defined in a total variability space was introduced. The total variability space representation is motivated by the popular Joint Factor Analysis (JFA) approach, but does not require the complication of estimating separate speaker and channel spaces and has been shown to be less dependent on score normalization procedures, such as znorm and t-norm. In this paper, we introduce a modification to the Cosine Similarity that does not require explicit score normalization, relying instead on simple mean and covariance statistics from a collection of impostor speaker ivectors. By avoiding the complication of zand t-norm, the new approach further allows for application of a new unsupervised speaker adaptation technique to models defined in the ivector space. Experiments are conducted on the core condition of the NIST 2008 corpora, where, with adaptation, the new approach produces an equal error rate (EER) of 4.8% and min decision cost function (MinDCF) of 2.3% on all female speaker trials.
-
Odyssey - Unsupervised Speaker Adaptation based on the Cosine Similarity for Text-Independent Speaker Verification.
2020Co-Authors: Stephen Shum, Najim Dehak, Reda Dehak, James GlassAbstract:This paper proposes a new approach to unsupervised speaker adaptation inspired by the recent success of the factor analysisbased Total Variability Approach to text-independent speaker verification [1]. This approach effectively represents speaker variability in terms of low-dimensional total factor vectors and, when paired alongside the simplicity of Cosine Similarity scoring, allows for easy manipulation and efficient computation [2]. The development of our adaptation algorithm is motivated by the desire to have a robust method of setting an adaptation threshold, to minimize the amount of required computation for each adaptation update, and to simplify the associated score normalization procedures where possible. To address the final issue, we propose the Symmetric Normalization (S-norm) method, which takes advantage of the symmetry in Cosine Similarity scoring and achieves competitive performance to that of the ZT-norm while requiring fewer parameter calculations. In subsequent experiments, we also assess an attempt to replace the use of score normalization procedures altogether with a Normalized Cosine Similarity scoring function [3]. We evaluated the performance of our unsupervised speaker adaptation algorithm under various score normalization procedures on the 10sec-10sec and core conditions of the 2008 NIST SRE dataset. Using results without adaptation as our baseline, it was found that the proposed methods are consistent in successfully improving speaker verification performance to achieve state-of-the-art results.
-
new Cosine Similarity scorings to implement gender independent speaker verification
Conference of the International Speech Communication Association, 2013Co-Authors: Mohammed Senoussaoui, Patrick Kenny, Pierre Dumouchel, Najim DehakAbstract:This paper is a natural extension of our previous work on gender-independent speaker verification systems [1]. In a previous paper, we presented a solution to avoid using gender information in the Probabilistic Linear Discriminant Analysis (PLDA) without any loss of accuracy compared with a genderdependent base-line implementation. In this work, we propose two solutions to make a speaker verification system based on Cosine Similarity independent of speaker gender. Our choice of the Cosine Similarity is motivated by the fact that it is proved itself as a second state-of-the art - in parallel with PLDA- of i-vector based speaker verification systems. As measured by Equal Error Rate and min DCF’s, performance results on the extended telephone list coreext-coreext condition of SRE2010 1 show no performance decrease in
-
Cosine Similarity scoring without score normalization techniques
Odyssey, 2010Co-Authors: Najim Dehak, Reda Dehak, James Glass, Douglas A Reynolds, Patrick KennyAbstract:In recent work [1], a simplified and highly effective approach to speaker recognition based on the Cosine Similarity between lowdimensional vectors, termed ivectors, defined in a total variability space was introduced. The total variability space representation is motivated by the popular Joint Factor Analysis (JFA) approach, but does not require the complication of estimating separate speaker and channel spaces and has been shown to be less dependent on score normalization procedures, such as znorm and t-norm. In this paper, we introduce a modification to the Cosine Similarity that does not require explicit score normalization, relying instead on simple mean and covariance statistics from a collection of impostor speaker ivectors. By avoiding the complication of zand t-norm, the new approach further allows for application of a new unsupervised speaker adaptation technique to models defined in the ivector space. Experiments are conducted on the core condition of the NIST 2008 corpora, where, with adaptation, the new approach produces an equal error rate (EER) of 4.8% and min decision cost function (MinDCF) of 2.3% on all female speaker trials.
-
unsupervised speaker adaptation based on the Cosine Similarity for text independent speaker verification
Odyssey, 2010Co-Authors: Stephen Shum, Najim Dehak, Reda Dehak, James GlassAbstract:This paper proposes a new approach to unsupervised speaker adaptation inspired by the recent success of the factor analysisbased Total Variability Approach to text-independent speaker verification [1]. This approach effectively represents speaker variability in terms of low-dimensional total factor vectors and, when paired alongside the simplicity of Cosine Similarity scoring, allows for easy manipulation and efficient computation [2]. The development of our adaptation algorithm is motivated by the desire to have a robust method of setting an adaptation threshold, to minimize the amount of required computation for each adaptation update, and to simplify the associated score normalization procedures where possible. To address the final issue, we propose the Symmetric Normalization (S-norm) method, which takes advantage of the symmetry in Cosine Similarity scoring and achieves competitive performance to that of the ZT-norm while requiring fewer parameter calculations. In subsequent experiments, we also assess an attempt to replace the use of score normalization procedures altogether with a Normalized Cosine Similarity scoring function [3]. We evaluated the performance of our unsupervised speaker adaptation algorithm under various score normalization procedures on the 10sec-10sec and core conditions of the 2008 NIST SRE dataset. Using results without adaptation as our baseline, it was found that the proposed methods are consistent in successfully improving speaker verification performance to achieve state-of-the-art results.
Jun Ye - One of the best experts on this subject based on the ideXlab platform.
-
Multicriteria Decision-Making Method Based On Cosine Similarity Measures Between Interval- Valued Fuzzy Sets With Risk Preference
Economic Computation and Economic Cybernetics Studies and Research, 2020Co-Authors: Jun YeAbstract:This paper presents the Cosine Similarity measure between IVFSs with risk preference and gives its decision making method using the Cosine Similarity measure depending on decision makers’ optimistic, neutral, and pessimistic natures for the subjective judgments that accompany the decision making process. Through the weighted Cosine Similarity measure between an alternative and the ideal alternative corresponding to one of optimistic, neutral, and pessimistic choices desired by decision makers, we can determine the ranking order of alternatives and the best one. This choosing feature corresponding to decision makers’ preference makes the proposed method not only more flexible, but also more suitable for many practical applications. Finally, an illustrative example is presented to demonstrate the feasibility and applicability of the proposed method.
-
Generalized Ordered Weighted Simplified Neutrosophic Cosine Similarity Measure for Multiple Attribute Group Decision Making
International Journal of Cognitive Informatics and Natural Intelligence, 2020Co-Authors: Jun YeAbstract:The paper proposes a generalized ordered weighted simplified neutrosophic Cosine Similarity (GOWSNCS) measure by combining the Cosine Similarity measure of simplified neutrosophic sets (SNSs) with the generalized ordered weighted averaging (GOWA) operator and investigates its properties and special cases. Then, the author develops a simplified neutrosophic group decision-making method based on the GOWSNCS measure to handle multiple attribute group decision-making problems with simplified neutrosophic information. The prominent characteristics of the GOWSNCS measure are that it not only is a generalization of the Cosine Similarity measure but also considers the associated weights for attributes and decision makers in the aggregation of the Cosine Similarity measures of SNSs to alleviate the influence of unduly large or small similarities in the process of information aggregation. Finally, an illustrative example of investment alternatives is provided to demonstrate the application and effectiveness of the developed approach.
-
improved Cosine Similarity measures of simplified neutrosophic sets for medical diagnoses
2015Co-Authors: Jun YeAbstract:Simplified neutrosophic set Single valued neutrosophic set Interval neutrosophic set Cosine Similarity measure Medical diagnosis a b s t r a c t Objective: In pattern recognition and medical diagnosis, Similarity measure is an important mathematical tool. To overcome some disadvantages of existing Cosine Similarity measures of simplified neutrosophic sets (SNSs) in vector space, this paper proposed improved Cosine Similarity measures of SNSs based on Cosine function, including single valued neutrosophic Cosine Similarity measures and interval neutrosophic Cosine Similarity measures. Then, weighted Cosine Similarity measures of SNSs were introduced by taking into account the importance of each element. Further, a medical diagnosis method using the improved Cosine Similarity measures was proposed to solve medical diagnosis problems with simplified neutrosophic information. Materials and methods: The improved Cosine Similarity measures between SNSs were introduced based on Cosine function. Then, we compared the improved Cosine Similarity measures of SNSs with existing Cosine Similarity measures of SNSs by numerical examples to demonstrate their effectiveness and rationality for overcoming some shortcomings of existing Cosine Similarity measures of SNSs in some cases. In the medical diagnosis method, we can find a proper diagnosis by the Cosine Similarity measures between the symptoms and considered diseases which are represented by SNSs. Then, the medical diagnosis method based on the improved Cosine Similarity measures was applied to two medical diagnosis problems to show the applications and effectiveness of the proposed method. Results: Two numerical examples all demonstrated that the improved Cosine Similarity measures of SNSs based on the Cosine function can overcome the shortcomings of the existing Cosine Similarity measures between two vectors in some cases. By two medical diagnoses problems, the medical diagnoses using various Similarity measures of SNSs indicated the identical diagnosis results and demonstrated the effectiveness and rationality of the diagnosis method proposed in this paper. Conclusions: The improved Cosine measures of SNSs based on Cosine function can overcome some drawbacks of existing Cosine Similarity measures of SNSs in vector space, and then their diagnosis method is very suitable for handling the medical diagnosis problems with simplified neutrosophic information and demonstrates the effectiveness and rationality of medical diagnoses.
-
improved Cosine Similarity measures of simplified neutrosophic sets for medical diagnoses
Artificial Intelligence in Medicine, 2015Co-Authors: Jun YeAbstract:We proposed improved Cosine Similarity measures of simplified neutrosophic sets (SNSs) based on Cosine function, including single valued neutrosophic Cosine Similarity measures and interval neutrosophic Cosine Similarity measures, to overcome some disadvantages of existing Cosine Similarity measures of SNSs.We presented a medical diagnosis method based on the improved Cosine Similarity measures to solve medical diagnosis problems with simplified neutrosophic information.Two medical diagnosis problems were given to show the effectiveness and rationality of the diagnosis method using the improved Cosine Similarity measures. ObjectiveIn pattern recognition and medical diagnosis, Similarity measure is an important mathematical tool. To overcome some disadvantages of existing Cosine Similarity measures of simplified neutrosophic sets (SNSs) in vector space, this paper proposed improved Cosine Similarity measures of SNSs based on Cosine function, including single valued neutrosophic Cosine Similarity measures and interval neutrosophic Cosine Similarity measures. Then, weighted Cosine Similarity measures of SNSs were introduced by taking into account the importance of each element. Further, a medical diagnosis method using the improved Cosine Similarity measures was proposed to solve medical diagnosis problems with simplified neutrosophic information. Materials and methodsThe improved Cosine Similarity measures between SNSs were introduced based on Cosine function. Then, we compared the improved Cosine Similarity measures of SNSs with existing Cosine Similarity measures of SNSs by numerical examples to demonstrate their effectiveness and rationality for overcoming some shortcomings of existing Cosine Similarity measures of SNSs in some cases. In the medical diagnosis method, we can find a proper diagnosis by the Cosine Similarity measures between the symptoms and considered diseases which are represented by SNSs. Then, the medical diagnosis method based on the improved Cosine Similarity measures was applied to two medical diagnosis problems to show the applications and effectiveness of the proposed method. ResultsTwo numerical examples all demonstrated that the improved Cosine Similarity measures of SNSs based on the Cosine function can overcome the shortcomings of the existing Cosine Similarity measures between two vectors in some cases. By two medical diagnoses problems, the medical diagnoses using various Similarity measures of SNSs indicated the identical diagnosis results and demonstrated the effectiveness and rationality of the diagnosis method proposed in this paper. ConclusionsThe improved Cosine measures of SNSs based on Cosine function can overcome some drawbacks of existing Cosine Similarity measures of SNSs in vector space, and then their diagnosis method is very suitable for handling the medical diagnosis problems with simplified neutrosophic information and demonstrates the effectiveness and rationality of medical diagnoses.
-
Multicriteria Decision-making Method using Cosine Similarity Measures for Reduct Fuzzy Sets of Interval-valued Fuzzy Sets
Journal of Computers, 2014Co-Authors: Jun YeAbstract:This paper introduces optimistic, neutral and pessimistic reduct fuzzy sets of an interval-valued fuzzy set, optimistic, neutral and pessimistic Cosine Similarity measures for the reduct fuzzy sets. A new decision-making method is proposed by means of three weighted Cosine Similarity measures depending on optimistic, neutral, and pessimistic points to reduce cognitive dissonance in multiple criteria decision analysis. We give the measures of optimism, neutralism, and pessimism to further determine suitability for alternative rankings through choosing optimistic, neutral, and pessimistic weighted Cosine Similarity measures. Finally, an illustrative example is conducted to validate the feasibility and applicability of the proposed method.
Hieu V Nguyen - One of the best experts on this subject based on the ideXlab platform.
-
Cosine Similarity metric learning for face verification
Asian Conference on Computer Vision, 2010Co-Authors: Hieu V NguyenAbstract:Face verification is the task of deciding by analyzing face images, whether a person is who he/she claims to be. This is very challenging due to image variations in lighting, pose, facial expression, and age. The task boils down to computing the distance between two face vectors. As such, appropriate distance metrics are essential for face verification accuracy. In this paper we propose a new method, named the Cosine Similarity Metric Learning (CSML) for learning a distance metric for facial verification. The use of Cosine Similarity in our method leads to an effective learning algorithm which can improve the generalization ability of any given metric. Our method is tested on the state-of-the-art dataset, the Labeled Faces in the Wild (LFW), and has achieved the highest accuracy in the literature.
-
ACCV (2) - Cosine Similarity metric learning for face verification
Computer Vision – ACCV 2010, 2010Co-Authors: Hieu V NguyenAbstract:Face verification is the task of deciding by analyzing face images, whether a person is who he/she claims to be. This is very challenging due to image variations in lighting, pose, facial expression, and age. The task boils down to computing the distance between two face vectors. As such, appropriate distance metrics are essential for face verification accuracy. In this paper we propose a new method, named the Cosine Similarity Metric Learning (CSML) for learning a distance metric for facial verification. The use of Cosine Similarity in our method leads to an effective learning algorithm which can improve the generalization ability of any given metric. Our method is tested on the state-of-the-art dataset, the Labeled Faces in the Wild (LFW), and has achieved the highest accuracy in the literature.
James Glass - One of the best experts on this subject based on the ideXlab platform.
-
Odyssey - Cosine Similarity Scoring without Score Normalization Techniques.
2020Co-Authors: Najim Dehak, Reda Dehak, James Glass, Douglas A Reynolds, Patrick KennyAbstract:In recent work [1], a simplified and highly effective approach to speaker recognition based on the Cosine Similarity between lowdimensional vectors, termed ivectors, defined in a total variability space was introduced. The total variability space representation is motivated by the popular Joint Factor Analysis (JFA) approach, but does not require the complication of estimating separate speaker and channel spaces and has been shown to be less dependent on score normalization procedures, such as znorm and t-norm. In this paper, we introduce a modification to the Cosine Similarity that does not require explicit score normalization, relying instead on simple mean and covariance statistics from a collection of impostor speaker ivectors. By avoiding the complication of zand t-norm, the new approach further allows for application of a new unsupervised speaker adaptation technique to models defined in the ivector space. Experiments are conducted on the core condition of the NIST 2008 corpora, where, with adaptation, the new approach produces an equal error rate (EER) of 4.8% and min decision cost function (MinDCF) of 2.3% on all female speaker trials.
-
Odyssey - Unsupervised Speaker Adaptation based on the Cosine Similarity for Text-Independent Speaker Verification.
2020Co-Authors: Stephen Shum, Najim Dehak, Reda Dehak, James GlassAbstract:This paper proposes a new approach to unsupervised speaker adaptation inspired by the recent success of the factor analysisbased Total Variability Approach to text-independent speaker verification [1]. This approach effectively represents speaker variability in terms of low-dimensional total factor vectors and, when paired alongside the simplicity of Cosine Similarity scoring, allows for easy manipulation and efficient computation [2]. The development of our adaptation algorithm is motivated by the desire to have a robust method of setting an adaptation threshold, to minimize the amount of required computation for each adaptation update, and to simplify the associated score normalization procedures where possible. To address the final issue, we propose the Symmetric Normalization (S-norm) method, which takes advantage of the symmetry in Cosine Similarity scoring and achieves competitive performance to that of the ZT-norm while requiring fewer parameter calculations. In subsequent experiments, we also assess an attempt to replace the use of score normalization procedures altogether with a Normalized Cosine Similarity scoring function [3]. We evaluated the performance of our unsupervised speaker adaptation algorithm under various score normalization procedures on the 10sec-10sec and core conditions of the 2008 NIST SRE dataset. Using results without adaptation as our baseline, it was found that the proposed methods are consistent in successfully improving speaker verification performance to achieve state-of-the-art results.
-
Cosine Similarity scoring without score normalization techniques
Odyssey, 2010Co-Authors: Najim Dehak, Reda Dehak, James Glass, Douglas A Reynolds, Patrick KennyAbstract:In recent work [1], a simplified and highly effective approach to speaker recognition based on the Cosine Similarity between lowdimensional vectors, termed ivectors, defined in a total variability space was introduced. The total variability space representation is motivated by the popular Joint Factor Analysis (JFA) approach, but does not require the complication of estimating separate speaker and channel spaces and has been shown to be less dependent on score normalization procedures, such as znorm and t-norm. In this paper, we introduce a modification to the Cosine Similarity that does not require explicit score normalization, relying instead on simple mean and covariance statistics from a collection of impostor speaker ivectors. By avoiding the complication of zand t-norm, the new approach further allows for application of a new unsupervised speaker adaptation technique to models defined in the ivector space. Experiments are conducted on the core condition of the NIST 2008 corpora, where, with adaptation, the new approach produces an equal error rate (EER) of 4.8% and min decision cost function (MinDCF) of 2.3% on all female speaker trials.
-
unsupervised speaker adaptation based on the Cosine Similarity for text independent speaker verification
Odyssey, 2010Co-Authors: Stephen Shum, Najim Dehak, Reda Dehak, James GlassAbstract:This paper proposes a new approach to unsupervised speaker adaptation inspired by the recent success of the factor analysisbased Total Variability Approach to text-independent speaker verification [1]. This approach effectively represents speaker variability in terms of low-dimensional total factor vectors and, when paired alongside the simplicity of Cosine Similarity scoring, allows for easy manipulation and efficient computation [2]. The development of our adaptation algorithm is motivated by the desire to have a robust method of setting an adaptation threshold, to minimize the amount of required computation for each adaptation update, and to simplify the associated score normalization procedures where possible. To address the final issue, we propose the Symmetric Normalization (S-norm) method, which takes advantage of the symmetry in Cosine Similarity scoring and achieves competitive performance to that of the ZT-norm while requiring fewer parameter calculations. In subsequent experiments, we also assess an attempt to replace the use of score normalization procedures altogether with a Normalized Cosine Similarity scoring function [3]. We evaluated the performance of our unsupervised speaker adaptation algorithm under various score normalization procedures on the 10sec-10sec and core conditions of the 2008 NIST SRE dataset. Using results without adaptation as our baseline, it was found that the proposed methods are consistent in successfully improving speaker verification performance to achieve state-of-the-art results.
Mohammed Senoussaoui - One of the best experts on this subject based on the ideXlab platform.
-
new Cosine Similarity scorings to implement gender independent speaker verification
Conference of the International Speech Communication Association, 2013Co-Authors: Mohammed Senoussaoui, Patrick Kenny, Pierre Dumouchel, Najim DehakAbstract:This paper is a natural extension of our previous work on gender-independent speaker verification systems [1]. In a previous paper, we presented a solution to avoid using gender information in the Probabilistic Linear Discriminant Analysis (PLDA) without any loss of accuracy compared with a genderdependent base-line implementation. In this work, we propose two solutions to make a speaker verification system based on Cosine Similarity independent of speaker gender. Our choice of the Cosine Similarity is motivated by the fact that it is proved itself as a second state-of-the art - in parallel with PLDA- of i-vector based speaker verification systems. As measured by Equal Error Rate and min DCF’s, performance results on the extended telephone list coreext-coreext condition of SRE2010 1 show no performance decrease in
