An information-theoretic perspective on feature selection in speaker recognition

Eriksson, Thomas; Kim, Sun Ah; Kang, Hong-Goo; Lee, Chungyong

doi:10.1109/lsp.2005.849495

Cited by 39 publications

(29 citation statements)

References 15 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The initial difference is made from the entropy definitions, such as Shannon entropy in [12,14,25,26], and Rényi entropy in [6,7,15]. The second difference is the selection of bound relations, such as "P e vs. H(X|Y)" in [12,24], "H(X|Y) vs. P e " in [6,7,14,15,20], "P e vs. MI(X; Y)" in [27,28], and "N MI(X; Y) vs. A" in [25], where A is the accuracy rate, MI(X; Y) and N MI(X; Y) are the mutual information and normalized mutual information between variables X and Y, respectively. Another important study is made on the tightness of bounds.…”

Section: Related Workmentioning

confidence: 99%

An Optimization Approach of Deriving Bounds between Entropy and Error from Joint Distribution: Case Study for Binary Classifications

Xing

2016

Entropy

View full text Add to dashboard Cite

Abstract:In this work, we propose a new approach of deriving the bounds between entropy and error from a joint distribution through an optimization means. The specific case study is given on binary classifications. Two basic types of classification errors are investigated, namely, the Bayesian and non-Bayesian errors. The consideration of non-Bayesian errors is due to the facts that most classifiers result in non-Bayesian solutions. For both types of errors, we derive the closed-form relations between each bound and error components. When Fano's lower bound in a diagram of "Error Probability vs. Conditional Entropy" is realized based on the approach, its interpretations are enlarged by including non-Bayesian errors and the two situations along with independent properties of the variables. A new upper bound for the Bayesian error is derived with respect to the minimum prior probability, which is generally tighter than Kovalevskij's upper bound.

show abstract

Section: Related Workmentioning

confidence: 99%

An Optimization Approach of Deriving Bounds between Entropy and Error from Joint Distribution: Case Study for Binary Classifications

Xing

2016

Entropy

View full text Add to dashboard Cite

show abstract

“…Eriksson et al 8 showed that the error rate of a speaker recognition system decreases as mutual information increases. In Eq.…”

Section: I͑c;x͒ = H͑c͒ − H͑c͉x͒ ͑1͒mentioning

confidence: 99%

“…In speaker recognition, some researchers use this measurement to measure or improve speaker recognition accuracy. 8,9 In this paper, we experimentally re-evaluate the speaker discriminative power of each phoneme class using mutual information and then find the optimal phoneme class ratio. We adopt the Nelder-Mead method, which is widely used for nonlinear optimization of multi-dimensional data.…”

Section: Introductionmentioning

confidence: 99%

Phonetically optimized speaker modeling for robust speaker recognition

Lee

Choi

Kang

2009

The Journal of the Acoustical Society of America

View full text Add to dashboard Cite

This paper proposes an efficient method to improve speaker recognition performance by dynamically controlling the ratio of phoneme class information. It utilizes the fact that each phoneme contains different amounts of speaker discriminative information that can be measured by mutual information. After classifying phonemes into five classes, the optimal ratio of each class in both training and testing processes is adjusted using a non-linear optimization technique, i.e., the Nelder–Mead method. Speaker identification results verify that the proposed method achieves 18% improvement in terms of error rate compared to a baseline system.

show abstract

“…Apparently, according to this formula, the feature subset with high fitness value will be possible to lead to high classification accuracy and less complex classifier. The corresponding fitness function of evaluation criterion 2 is: fitness 3 …”

Section: Consistency and Otherness Of Evaluation Criteriamentioning

confidence: 99%

“…In many pattern classification applications, selection of the most characterizing features (or attributes) of the observed data (such as feature selection or variable selection, among many other names) is important to maximize the classification accuracy [1][2][3][4][5][6][7][8][9]. This is especially important when one Y. Li ( ) · X. Zeng College of Communication Engineering, Chongqing University, Chongqing 400030, China e-mail: lymcentor924924@gmail.com is required to deal with a large or even overwhelming feature set.…”

Section: Introductionmentioning

confidence: 99%

Sequential multi-criteria feature selection algorithm based on agent genetic algorithm

Zeng

2008

Appl Intell

View full text Add to dashboard Cite

A multi-criteria feature selection method-sequential multi-criteria feature selection algorithm (SMCFS) has been proposed for the applications with high precision and low time cost. By combining the consistency and otherness of different evaluation criteria, the SMCFS adopts more than one evaluation criteria sequentially to improve the efficiency of feature selection. With one novel agent genetic algorithm (chain-like agent GA), the SMCFS can obtain high precision of feature selection and low time cost that is similar as filter method with single evaluation criterion. Several groups of experiments are carried out for comparison to demonstrate the performance of SMCFS. SMCFS is compared with different feature selection methods using three datasets from UCI database. The experimental results show that the SM-CFS can get low time cost and high precision of feature selection, and is very suitable for this kind of applications of feature selection.

show abstract

An information-theoretic perspective on feature selection in speaker recognition

Cited by 39 publications

References 15 publications

An Optimization Approach of Deriving Bounds between Entropy and Error from Joint Distribution: Case Study for Binary Classifications

An Optimization Approach of Deriving Bounds between Entropy and Error from Joint Distribution: Case Study for Binary Classifications

Phonetically optimized speaker modeling for robust speaker recognition

Sequential multi-criteria feature selection algorithm based on agent genetic algorithm

Contact Info

Product

Resources

About