Haydar Ankışhan

Square root unscented based FastSlam optimized by particle swarm optimization passive congregation

Simultaneous localization and mapping (SLAM) is known to be a problem for autonomous vehicles/robots. Different solutions have recently been proposed on this subject. The best known of these are FastSlam based approaches. In this study, two improved FastSlam based methods are proposed to solve the SLAM problem. In the first method, square root unscented (Sru) Kalman filter is used instead of extended Kalman filter in robot position prediction/update for each particle filter samples and feature updates. The second method uses Sru - Kalman filter with particle swarm optimization passive congregation (PSO-PC) for robot/feature position estimations. In the second method, particle swarm optimization passive congregation (PSO-PC) is used to optimize particle samples in case of sampling stage. The experimental results were compared with FastSlamII and unscented U-FastSlam. It is seen that proposed methods are an alternative for the solution of SLAM problem. The best results were obtained by Sru - based PSO-PC optimized FastSlam approach for the vehicle position and heading angle mean square errors.

Snore-related sound classification based on time-domain features by using ANFIS model

Obstructive sleep apnea/hypopnea (OSAH) is a highly prevalent disease which causes collapse in upper airway while sleeping. The purpose of this study is to classify snore related sounds into snore/non-snore episodes using adaptive neuro fuzzy inference system (ANFIS). Time-domain features which are entropy, energy and zero crossing rates were used and applied to data for ANFIS classifier model. At first, apnea and normal snore related sounds obtained from different patients are segmented. After segmentation, energy, entropy and zero crossing rates are calculated. Unlike the previous studies, entropy information was firstly used for snoring classification. Then, ANFIS was used to classify episodes as snore/non-snore. Experimental results have shown that ANFIS is able to classify snore segments with accuracy rate 97.08%. In conclusion, the results prove that ANFIS has good performance for classifying snore related sounds.

Square root unscented filter based FastSLAM approach for SLAM problem solution

There are different Bayesian based approaches proposed for the solution of simultaneous localization and mapping (SLAM) problem in the literature. In this study, square root unscented Kalman based (Sru)-FastSLAM and square root unscented particle filter based (SruPf) - FastSLAM were proposed for the SLAM problem solution. The first method used Sru - Kalman filter for estimating the robot position, the landmarks location and particle weights. The second method with the help of FastSlam II uses Sru-Kalman filter for each particle. FastSLAM II, unscented particle filter based (Upf) FastSlam II, unscented (U) FastSlam, unscented Kalman aided (UAided) FastSLAM, Sru- FastSlam and SruPf - FastSLAM were used for comparison of filter performance in the experimental results. It is seen that Sru - FastSlam and SruPf-FastSLAM are alternative to solving the problem of SLAM. The best results for heading, position error of robot/ vehicle and uncertainty of position of landmarks were obtained by Sru-FastSlam II.

Chaotic analysis of snore related sounds

Snoring is a noise related sound, which occurs while sleeping owing to narrow upper airway cause hard to breathing. There are a lot of studies available related to diagnosis and specifying the how harmful for people bodies. Some studies have used these sounds depending on acoustical features; the others have used polysomnography devices that produce some data to analyze the illness. In this study, we have used largest Lyapunov exponents (LLE) for chaotic analysis of snore related sounds in case of different snoring stage on sleeping. Estimated feature values were used for classifying of these sounds with adaptive neuro fuzzy inference system (ANFIS) classifier. These estimated feature values was used input for ANFIS classifier. ANFIS can also classify these sounds with the highest accuracy of 97% training and 89.08% testing results.

Slot Parameter Optimization for Multiband Antenna Performance Improvement Using Intelligent Systems

This paper discusses bandwidth enhancement for multiband microstrip patch antennas (MMPAs) using symmetrical rectangular/square slots etched on the patch and the substrate properties. The slot parameters on MMPA are modeled using soft computing technique of artificial neural networks (ANN). To achieve the best ANN performance, Particle Swarm Optimization (PSO) and Differential Evolution (DE) are applied with ANN’s conventional training algorithm in optimization of the modeling performance. In this study, the slot parameters are assumed as slot distance to the radiating patch edge, slot width, and length. Bandwidth enhancement is applied to a formerly designed MMPA fed by a microstrip transmission line attached to the center pin of 50 ohm SMA connecter. The simulated antennas are fabricated and measured. Measurement results are utilized for training the artificial intelligence models. The ANN provides 98% model accuracy for rectangular slots and 97% for square slots; however, ANFIS offer 90% accuracy with lack of resonance frequency tracking.

Adaptive neuro fuzzy supported Kalman filter approach for simultaneous localization and mapping

Simultaneous Localization and Mapping (SLAM) is a method employed by robots and autonomous vehicles to build up a map within an unknown environment or to update a map within a known environment. In recent years, SLAM has been a significant problem with autonomous. There have been different statistical methods used for solving this problem ranging from expectation maximization method to Kalman based filters and particle filters. In this study, square root uncented Kalman filter has been utilized to address the SLAM problem. Two basic improvements have been achieved with the proposed method i) tuning Q and R design matrices using adaptive neuro fuzzy inference system (ANFIS), ii) Rauch-Tung-Striebel smoother for enhancing the filters prediction. Simulation results have shown that the proposed filter is more successful compared with the extended, unscented, square root uncented Kalman filters and particle based FASTSLAM II model.

Analysing of the snore sound signals with AutoRegressive modelling

Obstructive sleep apnea (OSA) is a highly prevalent disease in which upper airways are collapsed during sleep, leading to serious consequences. The aim of this work is to study apnea, hypopnea and normal snoring sounds by using the criterias that are not used before in this area. The snoring sounds which are separated from segşments, that are in case of each inspiration and expiration, after enhanced by wavelet transform method. The AutoRegressive model order of these segments are determined with Final Prediction Error and Swartz Bayesion Criterion. Autocorrelation, Loss function and energy of segments are calculated on these sounds modelled with (AR) Autoregressive Model. The results were showed that the model order and energies of segments are the highest for patients of having apnea problem, middle degree for patients of having hypopnea problem and lowest degree for the patients of having normal snoring problems. In the meantime, loss function values were different for the patients of apnea, hypopnea and normal snoring patients. Data were obtained from Gulhane Military Medical Hospital at 20 patients. Those are 4 normal snoring, 8 hyponea problem and 8 apnea problem patients.

Dual Kalman filter approach for colored noise corrupted speech enhancement

In this paper, Kalman and Least Mean Square based filters are used for colored noise corrupted speech enhancement. Unlike previous studies a second speech signal has been utilized as colored noise which represents the situation where two persons are talking concurrently. Such a setup will help analyse the performance of speech enhancement algorithms when there are more than one speech components in the signal to be analysed and main speech signal has to be recovered. Final Prediction Error method has been employed for determining the model parameters, Speech was modeled with AR model and selected methods has been tested for their performance in terms of mean square error. The experimental results show that dual Kalman filter, which estimates both state and parameters concurently, has produced lower mean square error values when compared to joint and single Kalman filters. Joint Kalman filter, on the other hand, produced lower mean square error than single Kalman filter. Finally, it was observed that, the performance of LMS based filters was not adequate for the enhancement of colored noise corrupted speech.