Benyamin Kusumoputro

Fuzzy-Appearance Manifold and Fuzzy- Nearest Distance Calculation for Model- Less 3D Pose Estimation of Degraded Face Images

This paper presents the development of fuzzy-appearance manifold and fuzzy-nearest distance calculation in the eigenspace domain for pose estimation of degraded face images. In order to obtain a robust pose estimation system which can deal with the fuzziness of face data caused by statistical errors, we proposed the fuzzy-vector representation in eigenspace domain of the face images. Using fuzzy-vector representations, all of the crisp vectors of face data in the eigenspace domain are firstly transformed into fuzzy-vectors as fuzzy-points. Next, the fuzzy-appearance manifold is constructed from all the available fuzzy-points and the fuzzy-nearest distance calculation is proposed as the classifier of the pose estimation system. The pose estimation of an unknown face image is performed by firstly being projected onto the eigenspace domain then transformed to become an unknown fuzzy-point, and its fuzzy-distance with all of the available fuzzy-points in the fuzzy-appearance manifold will be calculated. The fuzzy-point in the manifold which has the nearest distance to that unknown fuzzy-point will be determined as the pose position of the unknown face image. In the experiment, face images with various quality degradation effects were used. The results show that the system could maintain high recognition rates for estimating the pose position of the degraded face images.

Improving recognition and generalization capability of back-propagation NN using a self-organized network inspired by immune algorithm (SONIA)

To improve recognition and generalization capability of back-propagation neural networks (BP-NN), a hidden layer self-organization inspired by immune algorithm called SONIA, is proposed. B cell construction mechanism of immune algorithm inspires a creation of hidden units having local data recognition ability that improves recognition capability. B cell mutation mechanism inspires a creation of hidden units having diverse data representation characteristics that improves generalization capability. Experiments on a sinusoidal benchmark problem show that the approximation error of the proposed network is 1/17 times lower than that of BP-NN. Experiments on real time-temperature-based food quality prediction data shows that the recognition capability is 18% improved comparing to that of BP-NN. The development of the world first time-temperature-based food quality prediction demonstrates the real applicability of the proposed method in the field of food industry.

Fuzzy-neuro LVQ and its comparison with fuzzy algorithm LVQ in artificial odor discrimination system.

The human sensory test is often used for obtaining the sensory quantities of odors, however, the fluctuation of results due to the experts condition can cause discrepancies among panelists. Authors have studied the artificial odor discrimination system using a quartz resonator sensor and a back-propagation neural network as the recognition system, however, the unknown category of odor is always recognized as the known category of odor. In this paper, a kind of fuzzy algorithm for learning vector quantization (LVQ) is developed and used as a pattern classifier. In this type of fuzzy LVQ, the neuron activation is derived through fuzziness of the input data, so that the neural system could deal with the statistics of the measurement error directly. During learning, the similarity between the training vector and the reference vectors are calculated, and the winning reference vector is updated by shifting the central position of the fuzzy reference vector toward or away from the input vector, and by modifying its fuzziness. Two types of fuzziness modifications are used, i.e., a constant modification factor and a variable modification factor. This type of fuzzy-neuro (FN) LVQ is different in nature from fuzzy algorithm (FA) LVQ, and in this paper, the performance of FNLVQ network is compared with that of FALVQ in an artificial odor recognition system. Experimental results show that both FALVQ and FNLVQ could provide high recognition probability in determining various known categories of odors, however, the FNLVQ neural system has the ability to recognize the unknown category of odor that could not be recognized by the FALVQ neural system.

Optimization of fuzzy-neural structure through genetic algorithms and its application in artificial odor recognition-system

Fuzzy neural networks are gaining much research interest and have attracted considerable attention recently, due to diverse applications in such fields as pattern recognition, image processing and control. However, this type of neural system, similar to that of multilayer perceptrons, has a drawback due to its huge neural connections. In this article, we proposed a method for optimizing the structure of a fuzzy artificial neural network (FANN) through genetic algorithms. This genetic algorithm (GA) is used to optimize the number of weight connections in a neural network structure, by evolutionary calculation of the fitness function of those structures as individuals in a population. The developed optimized fuzzy neural net is then applied for pattern recognition in an odor recognition system. Experimental results show that the optimized neural system provides higher recognition capability compared with that of unoptimized neural systems. The recognition rate of the unoptimized neural structure is 70.4% and could be increased to 85.2% in the optimized neural system. It is also shown that the computational cost of the optimized neural system structure is also lower than for the unoptimized structure.

Speaker identification in noisy environment using bispectrum analysis and probabilistic neural network

The paper describes the application of a neural processing for extracting bispectrum feature of speech data, and the use of probabilistic neural network as a classifier in an automatic speech recognition system. The usually used feature extraction paradigm in the early development of the speech recognition system is power spectrum analysis, however, the recognition rate of this system is not high enough, especially when a Gaussian noise is added to the utterance speech data. In this paper, we developed a speaker identification system using bispectrum feature analysis. To analyse the distribution of the bispectrum data along its two dimensional representation, we developed an adaptive feature extraction mechanism of the bispectrum speech data based on cascade neural network. A cascade configuration of SOFM (Self-Organizing Feature Map) and LVQ (Learning Vector Quantization) is used as an adaptive codebook generation algorithm for determining the feature distribution of the bispectrum speech data. The K-L transformation (K-LT) technique is then used as a preprocessing element before the neural classifier is utilized. This K-LT has shown as an effective procedure for orthogonalization and dimensionality reduction of the codebook vectors generated from bispectrum data. Experimental results show that our system could perform with high recognition rate on the undirected utterance speech, especially when a higher number of codebook vectors are utilized. It is also shown that the use of PNN could increase the recognition rate significantly, even using speech data with additional Gaussian noise.

A fuzzy-similarity-based self-organized network inspired by immune algorithm for three-mixture-fragrance recognition

A fuzzy-similarity-based self-organized network inspired by immune algorithm (F-SONIA) is proposed in order to develop an artificial odor discrimination system for three-mixture-fragrance recognition. It can deal with an uncertainty in frequency measurements, which is inherent in odor acquisition devices, by employing a fuzzy similarity. Mathematical analysis shows that the use of the fuzzy similarity results on a higher dissimilarity between fragrance classes, therefore, the recognition accuracy is improved and the learning time is reduced. Experiments show that F-SONIA improves recognition accuracy of SONIA by 3%-9% and the previously developed artificial odor discrimination system by 14%-25%. In addition, the learning time of F-SONIA is three times faster than that of SONIA.

Improvement of artificial odor discrimination system using fuzzy-LVQ neural network

An artificial odor recognition system is developed in order to mimic the human sensory test in cosmetics, perfume and beverage industries. A backpropagation neural network is used as the pattern recognition system and shows high recognition capability. However, the system only works efficiently when it is used to discriminate a limited number of odors. The unlearned odor will be classified as one of the already learned category. To improve the systems capability, a fuzzy learning vector quantization neural network is developed and utilized in experiments on four different ethanol concentrations, and three different kinds of fragrance odor from Martha Tilaar Cosmetics. The results shows that the FLVQ has a comparable ability for recognizing the already known category of odors. However, the FLVQ algorithm can cluster the unknown odor in a different new class of odor.

Improving the artificial odor and gas source localization system using the semiconductor gas sensor based on RF communication

Locating gas leaks, chemical hazard sources and smuggled narcotics are difficult tasks to be done by conventional systems, because they usually exploit human experts or dog trackers to search and solve the problems. An electronic system that can perform such tracking automatically, constantly and accurately is expected to replace these duties. This paper describes how an artificial system with such ability is realized. It can be used as a tool to conduct various applications, such as locating gas leaks and searching for chemical hazards.

Identification of Industrial Furnace Temperature for Sintering Process in Nuclear Fuel Fabrication Using NARX Neural Networks

Nonlinear system identification is becoming an important tool which can be used to improve control performance and achieve robust fault-tolerant behavior. Among the different nonlinear identification techniques, methods based on neural network model are gradually becoming established not only in the academia, but also in industrial application. An identification scheme of nonlinear systems for sintering furnace temperature in nuclear fuel fabrication using neural network autoregressive with exogenous inputs (NNARX) model investigated in this paper. The main contribution of this paper is to identify the appropriate model and structure to be applied in control temperature in the sintering process in nuclear fuel fabrication, that is, a nonlinear dynamical system. Satisfactory agreement between identified and experimental data is found with normalized sum square error 1 for heating step and for soaking step. That result shows the model successfully predict the evolution of the temperature in the furnace.

Localizing multiple odor sources in dynamic environment using ranged subgroup PSO with flow of wind based on open dynamic engine library

A new algorithm based on Modified Particle Swarm Optimization (MPSO) which follows a local gradient of the chemical concentration within a plume and follow direction of the wind velocity is investigated. Moreover, the niche or parallel search characteristic is adopted on MPSO to solve the multi-peak and multi-source problem. When using parallel MPSO, subgroup of robot is introduced then each subgroup can locate the odor source. Unfortunately, there is a possibility that more that one subgroup locates one odor sources. This is inefficient because other subgroups locate other source, then we proposed a ranged subgroup method for coping for that problem, then the searching performance will increase. Then ODE (Open Dynamics Engine) library is used for physical modeling of the robot like friction, balancing moment and others. Finally the statistical analysis shows that the new approach is technically sounds.