• Proceedings of the KIEE Conference
• /
• 2008.04a
• /
• pp.99-100
• /
• 2008

This paper proposes a Gaussian Mixture Model (GMM) - based speech emotion recognition methods using four feature parameters; 1) Fast Fourier Transform(FFT) spectral entropy, 2) delta FFT spectral entropy, 3) Mel-frequency Filter Bank (MFB) spectral entropy, and 4) delta MFB spectral entropy. In addition, we use four emotions in a speech database including anger, sadness, happiness, and neutrality. We perform speech emotion recognition experiments using each pre-defined emotion and gender. The experimental results show that the proposed emotion recognition using FFT spectral-based entropy and MFB spectral-based entropy performs better than existing emotion recognition based on GMM using energy, Zero Crossing Rate (ZCR), Linear Prediction Coefficient (LPC), and pitch parameters. In experimental Results, we attained a maximum recognition rate of 75.1% when we used MFB spectral entropy and delta MFB spectral entropy.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.35 no.12C
• /
• pp.995-1003
• /
• 2010

The content-based hashing for authentication and copy protection of image, video and 3D model has to satisfy the robustness and the security. For the security analysis of the hash value, the modelling method based on differential entropy had been presented. But this modelling can be only applied to the image hashing. This paper presents the modelling for the security analysis of the hash feature value in 3D model hashing based on differential entropy. The proposed security analysis modeling design the feature extracting methods of two types and then analyze the security of two feature values by using differential entropy modelling. In our experiment, we evaluated the security of feature extracting methods of two types and discussed about the trade-off relation of the security and the robustness of hash value.

• Journal of Information Processing Systems
• /
• v.14 no.1
• /
• pp.218-227
• /
• 2018

Vocal effort detection is important for both robust speech recognition and speaker recognition. In this paper, the spectral information entropy feature which contains more salient information regarding the vocal effort level is firstly proposed. Then, the model fusion method based on complementary model is presented to recognize vocal effort level. Experiments are conducted on isolated words test set, and the results show the spectral information entropy has the best performance among the three kinds of features. Meanwhile, the recognition accuracy of all vocal effort levels reaches 81.6%. Thus, potential of the proposed method is demonstrated.

• Transactions of the Korean Society of Mechanical Engineers A
• /
• v.30 no.7 s.250
• /
• pp.787-793
• /
• 2006

Recently kriging model has been widely used in the DACE (Design and Analysis of Computer Experiment) because of prominent predictability of nonlinear response. Since DACE has no random or measurement errors contrast to physical experiment, space filling experimental design that distributes uniformly design points over whole design space should be employed as a sampling method. In this paper, we examine the maximum entropy experimental design that reveals the space filling strategy in which defines the maximum entropy based on Gaussian or exponential. The influence of these two correlation functions on space filling design and their model parameters are investigated. Based on the exploration of numerous numerical tests, enhanced maximum entropy design based on exponential correlation function is suggested.

• Journal of Information Processing Systems
• /
• v.20 no.3
• /
• pp.285-295
• /
• 2024

Driven by the vague assessment big data, a product service system (PSS) evaluation method is developed based on a hybrid model of multi-weight combination and improved TOPSIS by relative entropy. The index values of PSS alternatives are solved by the integration of the stakeholders' vague assessment comments presented in the form of trapezoidal fuzzy numbers. Multi-weight combination method is proposed for index weight solving of PSS evaluation decision-making. An improved TOPSIS by relative entropy (RE) is presented to overcome the shortcomings of traditional TOPSIS and related modified TOPSIS and then PSS alternatives are evaluated. A PSS evaluation case in a printer company is given to test and verify the proposed model. The RE closeness of seven PSS alternatives are 0.3940, 0.5147, 0.7913, 0.3719, 0.2403, 0.4959, and 0.6332 and the one with the highest RE closeness is selected as the best alternative. The results of comparison examples show that the presented model can compensate for the shortcomings of existing traditional methods.

• Communications for Statistical Applications and Methods
• /
• v.19 no.6
• /
• pp.837-847
• /
• 2012

Inverse Gaussian distribution is widely used in applications to analyze and model right-skewed data. To assess the appropriateness of the distribution prior to data analysis, Mudholkar and Tian (2002) proposed an entropy-based test of fit. The test is based on the entropy power fraction(EPF) index suggested by Gokhale (1983). The simulation results report that the power of the entropy-based test is superior compared to other goodness-of-fit tests; however, this observation is based on the small-scale simulation results on the standard exponential, Weibull W(1; 2) and lognormal LN(0:5; 1) distributions. A large-scale simulation should be performed against various alternative distributions to evaluate the power of the entropy-based test; however, the use of a theoretical method is more effective to investigate the powers. In this paper, utilizing the information discrimination(ID) index defined by Ehsan et al. (1995) as a mathematical tool, we scrutinize the power of the entropy-based test. The selected alternative distributions are the gamma, Weibull and lognormal distributions, which are widely used in data analysis as an alternative to inverse Gaussian distribution. The study results are provided and an illustrative example is analyzed.

• Journal of the Korean Society of Safety
• /
• v.38 no.1
• /
• pp.1-8
• /
• 2023

It is extremely important to design safety-guaranteed industrial processes because such process determine the ultimate outcomes of industrial activities, including worker safety. Application of artificial intelligence (AI) in industrial safety involves modeling industrial safety systems by using vast amounts of safety-related data, accident prediction, and accident prevention based on predictions. As a preliminary step toward realizing AI-based industrial safety in Korea, this study discusses neural network-based modeling of industrial safety systems. The input variables that are the most discriminatory relative to the output variables of industrial safety processes are selected using two information-theoretic measures, namely entropy and cross entropy. Normalized frequency and severity of industrial accidents are selected as the output variables. Our simulation results confirm the effectiveness of the proposed neural network model and, therefore, the feasibility of extending the model to include more input and output variables.

• Journal of Communications and Networks
• /
• v.5 no.1
• /
• pp.65-72
• /
• 2003

Internet roundtrip delay/time (RTT) prediction plays an important role in detecting packet losses in reliable transport protocols for traditional web applications and determining proper transmission rates in many rate-based TCP-friendly protocols for Internet-based real-time applications. The widely adopted autoregressive and moving average (ARMA) model with fixed-parameters is shown to be insufficient for all scenarios due to its intrinsic limitation that it filters out all high-frequency components of RTT dynamics. In this paper, we introduce a novel parameter-varying RTT model for Internet roundtrip time prediction based on the information theory and the maximum entropy principle (MEP). Since the coefficients of the proposed RTT model are updated dynamically, the model is adaptive and it tracks RTT dynamics rapidly. The results of our experiments show that the MEP algorithm works better than the ARMA method in both RTT prediction and RTO estimation.

• Bulletin of the Korean Chemical Society
• /
• v.29 no.5
• /
• pp.1051-1054
• /
• 2008

The entropy production in a non-equilibrium state based on the reversible Oregonator model of the Belousov-Zhabotinskii (BZ) reaction system has been studied. The reaction affinity and the reaction rate for the individual steps have been calculated by varying the concentrations of key variables in the system. The result shows a linear relationship between the reaction affinity and the reaction rate in the given concentration range. However, the overall entropy calculated on the basic assumption that the entropy in a reaction system corresponds to the summation of a product of reaction affinity and reaction rate of individual steps shows a nonlinearity of the reaction system. The results well agrees with the fact that the entropy production is not linear or complicated function in a non-linear reaction system.

• ETRI Journal
• /
• v.46 no.3
• /
• pp.513-525
• /
• 2024

Deep neural networks trained on labeled medical data face major challenges owing to the economic costs of data acquisition through expensive medical imaging devices, expert labor for data annotation, and large datasets to achieve optimal model performance. The heterogeneity of diseases, such as Alzheimer's disease, further complicates deep learning because the test cases may substantially differ from the training data, possibly increasing the rate of false positives. We propose a reconstruction-based self-supervised anomaly detection model to overcome these challenges. It has a dual-subnetwork encoder that enhances feature encoding augmented by skip connections to the decoder for improving the gradient flow. The novel encoder captures local and global features to improve image reconstruction. In addition, we introduce an entropy-based image conversion method. Extensive evaluations show that the proposed model outperforms benchmark models in anomaly detection and classification using an encoder. The supervised and unsupervised models show improved performances when trained with data preprocessed using the proposed image conversion method.