BazTech - Yadda

1

Fire Detection Methods Based on Various Color Spaces and Gaussian Mixture Models

Munshi Amr

Advances in Science and Technology. Research Journal

|

2021

|

Vol. 15, no 3

197--214

EN

Fire disasters are very serious problems that may cause damages to ecological systems, infrastructure, properties, and even a threat to human lives; therefore, detecting fires at their earliest stage is of importance. Inspired by the technological advancements in artificial intelligence and image processing in solving problems in different applications, this encourages adopting those technologies in reducing the damage and harm caused by fire. This study attempts to propose an intelligent fire detection method by investigating three approaches to detect fire based on three different color models: RGB, YCbCr, and HSV are presented. The RGB method is applied based on the relationship among the red, green and blue values of pixels in images. In the YCbCr color model, image processing and machine learning techniques are used for morphological processing and automatic recognition of fire images. Whereas for the HSV supervised machine learning techniques are adopted, namely decision rule and Gaussian mixture model (GMM). Further, the expectation maximization (EM) algorithm is deployed for the GMM parameters estimation. The three proposed models were tested on two data sets, one of which contains fire images, the other consists of non-fire images with some having fire-like colors to test the efficiency of the proposed methods. The experimental results showed that the overall accuracies on two data sets for the RGB, YCbCr, and HSV methods were satisfactory and were efficient in detecting outdoor and indoor fires.

2

Enhancement in Bearing Fault Classification Parameters Using Gaussian Mixture Models and Mel Frequency Cepstral Coefficients Features

Atmani Youcef, Rechak Said, Mesloub Ammar, Hemmouche Larbi

Archives of Acoustics

|

2020

|

Vol. 45, No. 2

283--295

EN

Last decades, rolling bearing faults assessment and their evolution with time have been receiving much interest due to their crucial role as part of the Conditional Based Maintenance (CBM) of rotating machinery. This paper investigates bearing faults diagnosis based on classification approach using Gaussian Mixture Model (GMM) and the Mel Frequency Cepstral Coefficients (MFCC) features. Throughout, only one criterion is defined for the evaluation of the performance during all the cycle of the classification process. This is the Average Classification Rate (ACR) obtained from the confusion matrix. In every test performed, the generated features vectors are considered along to discriminate between four fault conditions as normal bearings, bearings with inner and outer race faults and ball faults. Many configurations were tested in order to determinate the optimal values of input parameters, as the frame analysis length, the order of model, and others. The experimental application of the proposed method was based on vibration signals taken from the bearing datacenter website of Case Western Reserve University (CWRU). Results show that proposed method can reliably classify different fault conditions and have a highest classification performance under some conditions.

3

Speaker Model Clustering to Construct Background Models for Speaker Verification

Dişken G., Tüfekci Z., Çevik U.

Archives of Acoustics

|

2017

|

Vol. 42, No. 1

127--135

EN

Conventional speaker recognition systems use the Universal Background Model (UBM) as an imposter for all speakers. In this paper, speaker models are clustered to obtain better imposter model representations for speaker verification purpose. First, a UBM is trained, and speaker models are adapted from the UBM. Then, the k-means algorithm with the Euclidean distance measure is applied to the speaker models. The speakers are divided into two, three, four, and five clusters. The resulting cluster centers are used as background models of their respective speakers. Experiments showed that the proposed method consistently produced lower Equal Error Rates (EER) than the conventional UBM approach for 3, 10, and 30 seconds long test utterances, and also for channel mismatch conditions. The proposed method is also compared with the i-vector approach. The three-cluster model achieved the best performance with a 12.4% relative EER reduction in average, compared to the i-vector method. Statistical significance of the results are also given.

4

Biometric speech signal processing in a system with digital signal processor

Marciniak T., Weychan R., Stankiewicz A., Dąbrowski A.

Bulletin of the Polish Academy of Sciences. Technical Sciences

|

2014

|

Vol. 62, nr 3

589--594

EN

This paper presents an analysis of issues related to the fixed-point implementation of a speech signal applied to biometric purposes. For preparing the system for automatic speaker identification and for experimental tests we have used the Matlab computing environment and the development software for Texas Instruments digital signal processors, namely the Code Composer Studio (CCS). The tested speech signals have been processed with the TMS320C5515 processor. The paper examines limitations associated with operation of the realized embedded system, demonstrates advantages and disadvantages of the technique of automatic software conversion from Matlab to the CCS and shows the impact of the fixed-point representation on the speech identification effectiveness.

5

Estimation of parameters of Gaussian mixture models by a hybrid method combining a self-adaptive differential evolution with the EM algorithm

Kwedlo W.

Advances in Computer Science Research

|

2014

|

Nr 11

109--123

EN

In the paper the problem of learning of Gaussian mixture models (GMMs) is considered. A new approach based on hybridization of a self-adaptive version of differential evolution (DE) with the classical EM algorithm is described. In this approach, called DEEM, the EM algorithm is run until convergence to fine-tune each solution obtained by the mutation and crossover operators of DE. To avoid the problem with parameter representation and infeasible solutions we use a method in which the covariance matrices are encoded using their Cholesky factorizations. In a simulation study GMMs were used to cluster synthetic datasets differing by a degree of separation between clusters. The results of experiments indicate that DE-EM outperforms the standard multiple restart expectation-maximization algorithm (MREM). For datasets with high number of features it also outperforms the state of-the-art random swap EM (RSEM).

PL

W pracy poruszono problem uczenia modeli mieszanin rozkładów normalnych. Zaproponowano nowe podejście, nazwane DE-EM, oparte na hybrydyzacji samoadaptacyjnego algorytmu ewolucji różnicowej i klasycznego algorytmu EM. W nowej metodzie rozwiązanie otrzymane jako wynik operatorów mutacji i krzyżowania jest poddawane optymalizacji lokalnej, prowadzonej aż do momentu uzyskania zbieżności, przez algorytm EM. Aby uniknąć problemu z reprezentacją macierzy kowariancji i niedopuszczalności rozwiązań użyto metody, w której macierze kowariancji są kodowane przy pomocy dekompozycji Cholesky’ego. W badaniach symulacyjnych modele mieszanin rozkładów normalnych zastosowano do grupowania danych syntetycznych. Wyniki eksperymentów wskazują, że metoda DE-EM osiąga lepsze wyniki niż standardowa technika wielokrotnego startu algorytmu ˙ EM. Dla zbiorów danych z dużą liczbą cech, metoda osiąga lepsze wyniki niż technika losowej wymiany rozwiązań połączona z algorytmem EM.

6

Construction and verification of mathematical model of mass spectrometry data

Plechawska-Wójcik M.

Informatyka, Automatyka, Pomiary w Gospodarce i Ochronie Środowiska

|

2013

|

nr 1

9--14

EN

The article presents issues concerning construction, adjustment and implementation of mass spectrometry mathematical model based on Gaussians and Mixture Models and the mean spectrum. This task is essential to the analysis and it needs specification of many parameters of the model.

PL

Artykuł przedstawia kwestie związane z konstrukcją, dopasowaniem i implementacją modelu matematycznego widm masowych opartego o rozkłady normalne i mieszaniny rozkładów oraz o widmo średnie. To zadanie jest kluczowe dla analizy, wymaga też określenia wielu parametrów modelu.

7

Voice Conversion Based on Hybrid SVR and GMM

Song P., Jin Y., Zhao L., Zou C.

Archives of Acoustics

|

2012

|

Vol. 37, No. 2

143-149

EN

A novel VC (voice conversion) method based on hybrid SVR (support vector regression) and GMM (Gaussian mixture model) is presented in the paper, the mapping abilities of SVR and GMM are exploited to map the spectral features of the source speaker to those of target ones. A new strategy of F0 transfor- mation is also presented, the F0s are modeled with spectral features in a joint GMM and predicted from the converted spectral features using the SVR method. Subjective and objective tests are carried out to evaluate the VC performance; experimental results show that the converted speech using the proposed method can obtain a better quality than that using the state-of-the-art GMM method. Meanwhile, a VC method based on non-parallel data is also proposed, the speaker-specific information is investigated us- ing the SVR method and preliminary subjective experiments demonstrate that the proposed method is feasible when a parallel corpus is not available.

8

Hierarchical Classification of Environmental Noise Sources Considering the Acoustic Signature of Vehicle Pass-Bys

Valero X., Alias F.

Archives of Acoustics

|

2012

|

Vol. 37, No. 4

423-434

EN

This work is focused on the automatic recognition of environmental noise sources that affect humans’ health and quality of life, namely industrial, aircraft, railway and road traffic. However, the recognition of the latter, which have the largest influence on citizens’ daily lives, is still an open issue. Therefore, although considering all the aforementioned noise sources, this paper especially focuses on improving the recognition of road noise events by taking advantage of the perceived noise differences along the road vehicle pass-by (which may be divided into different phases: approaching, passing and receding). To that effect, a hierarchical classification scheme that considers these phases independently has been implemented. The proposed classification scheme yields an averaged classification accuracy of 92.5%, which is, in absolute terms, 3% higher than the baseline (a traditional flat classification scheme without hierarchical structure). In particular, it outperforms the baseline in the classification of light and heavy vehicles, yielding a classification accuracy 7% and 4% higher, respectively. Finally, listening tests are performed to compare the system performance with human recognition ability. The results reveal that, although an expert human listener can achieve higher recognition accuracy than the proposed system, the latter outperforms the non-trained listener in 10% in average.

9

Statystyczna analiza danych proteomicznych

Kawulok J., Polańska J.

Studia Informatica

|

2011

|

Vol. 32, nr 2A

229-243

PL

Celem prac przedstawionych w niniejszym artykule była konstrukcja narzędzi do analizy danych proteomicznych, które pomogą w wykrywaniu choroby nowotworowej we wczesnym jej stadium. W prezentowanym artykule na początku została przeprowadzona wstępna obróbka widm masowych. Po zamodelowaniu ich za pomocą mieszanin gaussowskich wykorzystano klasyfikator SVM do rozdzielenia grupy z rakiem od grupy kontrolnej. Przedstawione wyniki badań potwierdziły skuteczność wykonanych operacji.

EN

The aim of the work reported in this paper was to develop statistical tools for mass spectra analysis. They would make it possible to detect cancer at its early stages. The main goal was to construct a classifier which would best distinguish people with cancer from a control group. First, the mass spectral signal is pre-processed. Next, the signals are modeled using Gaussian mixtures and they are later classified. The obtained results confirmed the effectiveness of the presented method.

10

Designing Model Based Classifiers by Emphasizing Soft Targets

El Jelali S., Lyhyaoui A., Figueiras-Vidal A.R.

Fundamenta Informaticae

|

2009

|

Vol. 96, nr 4

419-433

EN

When training machine classifiers, to replace hard classification targets by emphasized soft versions of them helps to reduce the negative effects of using standard cost functions as approximations to misclassification rates. This emphasis has the same kind of effect as sample editing methods, that have proved to be effective for improving classifiers performance. In this paper, we explore the effectiveness of using emphasized soft targets with generative models, such as Gaussian MixtureModels (GMM), and Gaussian Processes (GP). The interest of using GMMis that they offer advantages such as an easy interpretation and straightforward possibilities to deal with missing values. With respect to GP, if we use soft targets, we do not need to resort to any complex approximation to get a Gaussian Process classifier and, simultaneously, we can obtain the advantages provided by the use of an emphasis. Simulation results support the usefulness of the proposed approach to get better performance and show a low sensitivity to design parameters selection.

11

Fault classification in cylinders using multi-layer perceptrons, support vector machines and Gaussian mixture models

Marwala T., Mahola U., Chakraverty S.

Computer Assisted Mechanics and Engineering Sciences

|

2007

|

Vol. 14, No. 2

307-316

EN

Gaussian mixture models (GMM) and support vector machines (SVM) are introduced to classify faults in a population of cylindrical shells. The proposed procedures are tested on a population of 20 cylindrical shells and their performance is compared to the procedure, which uses multi-layer perceptrons (MLP). The modal properties extracted from vibration data are reduced into low dimension using the principal component analysis and are then used to train the GMM, SVM and MLP. It is observed that the GMM gives 98% classification accuracy, SVM gives 94% classification accuracy while the MLP gives 88% classification accuracy. Furthermore, GMM is found to be more computationally efficient than MLP which is in turn more computationally efficient than SVM.