Feature selection using genetic algorithmand its application to speaker verification

Haydar, Ali; Demirekler, Mübeccel; Yurtseven, M.K.

doi:10.1049/el:19981069

Cited by 10 publications

(3 citation statements)

References 6 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This is because of the fact that the performance of classifier and the cost of classification are sensitive to the selection of the features used in the construction of the classifier. By reducing the set of features, the time required for learning the classification knowledge and the time needed for classification reduces [26]. Further, by the extraction of relevant features and therefore the elimination of the irrelevant ones, the accuracy of the classifier can be increased.…”

Section: Feature Set Selection Using Genetic Algorithmmentioning

confidence: 99%

Automatic ECG Beat Arrhythmia Detection

Bazarghan,

Jaberi,

Amandi

et al. 2012

Preprint

View full text Add to dashboard Cite

Background: In recent years automated data analysis techniques have drawn great attention and are used in almost every field of research including biomedical. Artificial Neural Networks (ANNs) are one of the Computer-Aided-Diagnosis tools which are used extensively by advances in computer hardware technology. The application of these techniques for disease diagnosis has made great progress and is widely used by physicians.An Electrocardiogram carries vital information about heart activity and physicians use this signal for cardiac disease diagnosis which was the great motivation towards our study.Methods: In this study we are using Probabilistic Neural Networks (PNN) as an automatic technique for ECG signal analysis along with a Genetic Algorithm (GA). As every real signal recorded by the equipment can have different artifacts, we need to do some preprocessing steps before feeding it to the ANN. Wavelet transform is used for extracting the morphological parameters and median filter for data reduction of the ECG signal. The subset of morphological parameters are chosen and optimized using GA. We had two approaches in our investigation, the first one uses the whole signal with 289 normalized and de-noised data points as input to the ANN. In the second approach after applying all the preprocessing steps the signal is reduced to 29 data points and also their important parameters extracted to form the ANN input with 35 data points.Results: The outcome of the two approaches for 8 types of arrhythmia shows that the second approach is superior than the first one with an average accuracy of %99.42 . Conclusions:We have studied 8 types of arrhythmia with high detection accuracy. In the literature, previous attempts are made on 6 types of arrhythmias. The results of the PNN shows that its performance for reduced input signal along with the morphological parameter has the best performance. Also it was noticed that the proper selection of training and testing data sets are of great importance and all the beats of each arrhythmia should not be selected from a single file in the database.

show abstract

Section: Feature Set Selection Using Genetic Algorithmmentioning

confidence: 99%

Automatic ECG Beat Arrhythmia Detection

Bazarghan,

Jaberi,

Amandi

et al. 2012

Preprint

View full text Add to dashboard Cite

show abstract

“…Research on speaker verification has been focused on speaker models [2], feature selection [3], and robust methods [4]. Higgins used a discriminate counting to verify the speaker [5], as well as likelihood score normalization methods [6]- [8], which are two likelihood score normalization methods by using impostor models.…”

Section: Introductionmentioning

confidence: 99%

“…GSMSV is different from both the speaker verification methods with the conventional likelihood score (noted as CSV method in the following) [3] and the normalization method proposed in [6]. As we know, CSV method has some limitations, i.e., the loose distribution of likelihood scores leading to the vague boundaries between speakers and the burden to set a proper threshold, as well as the low system adaptability to protean input utterances.…”

Section: Introductionmentioning

confidence: 99%

A novel text-independent speaker verification method based on the global speaker model

Zhang

Zhu

2000

IEEE Trans. Syst., Man, Cybern. A

View full text Add to dashboard Cite

This correspondence introduces a new text-independent speaker verification method, which is derived from the basic idea of pattern recognition that the discriminating ability of a classifier can be improved by removing the common information between classes. In looking for the common speech characteristics between a group of speakers, a global speaker model can be established. By subtracting the score acquired from this model, the conventional likelihood score is normalized with the consequence of more compact score distribution and lower equal error rates. Several experiments are carried out to demonstrate the effectiveness of the proposed method.

show abstract

Retracted: Feature Selection Using Ant Colony Optimization for Text-Independent Speaker Verification System

Sohafi-Bonab

Aghdam

2010

Advances in Computation and Intelligence

View full text Add to dashboard Cite

Feature selection using genetic algorithmand its application to speaker verification

Cited by 10 publications

References 6 publications

Automatic ECG Beat Arrhythmia Detection

Automatic ECG Beat Arrhythmia Detection

A novel text-independent speaker verification method based on the global speaker model

Retracted: Feature Selection Using Ant Colony Optimization for Text-Independent Speaker Verification System

Contact Info

Product

Resources

About