Moon Kyou Song

L2-L∞ Filtering for Takagi-Sugeno fuzzy neural networks based on Wirtinger-type inequalities

Abstract This paper deals with the L 2 – L ∞ filtering problem for continuous-time Takagi–Sugeno fuzzy delayed Hopfield neural networks based on Wirtinger-type inequalities. A new set of delay-dependent conditions is established to estimate the state variables of fuzzy neural networks through the observed input and output variables. This ensures that the state estimation error system is asymptotically stable with a guaranteed L 2 – L ∞ performance. The presented criterion is formulated in terms of linear matrix inequalities (LMIs). An example with simulation results is given to illustrate the effectiveness of the proposed fuzzy neural state estimator.

Performance analysis of cell search in W-CDMA systems over Rayleigh fading channels

The three-step cell search has been considered for fast acquisition of the scrambling code unique to a cell in wideband code-division multiple-access (W-CDMA) systems. In this paper, the performance of the cell-search scheme is analyzed in Rayleigh fading channels. The system parameters for the cell-search scheme and the design parameters for the receivers are examined. Probabilities of detection, false alarm, and miss for each of the three steps are derived in closed form based on the statistics of CDMA noncoherent demodulator output. Through the analysis, the effect of threshold setting and post-detection integration for each step is investigated and the power allocation for the channels is considered. The optimal number of post-detection integrations for each step may depend on not only the power allocation for the channels related to the cell search, but also the false-alarm penalty time. Our analysis can be utilized for determining the values of the parameters. Also, the cumulative probability distribution of the average cell-search time is obtained. Finally, it is shown that the analysis could be used to obtain the distribution of the cell-search time graphically according to the distance between a mobile station and a base station by considering the propagation models for pass loss and the traffic distribution models.

Fuzzy horizon group shift FIR filtering for nonlinear systems with TakagiSugeno model

In recent years, the TakagiSugeno (TS) fuzzy model has been commonly used for the approximation of nonlinear systems. Using the TS fuzzy model, nonlinear systems can be converted into linear time-varying systems, which can reduce approximation errors compared with the conventional Taylor approximation. In this paper, we propose a new nonlinear filter with a finite impulse response (FIR) structure based on the TS fuzzy model. We firstly derive the fuzzy FIR filter and combine it with the horizon group shift (HGS) algorithm to manage the horizon size, which is an important design parameter of FIR filters. The resulting filter is called the fuzzy HGS FIR filter (FHFF). Due to the FIR structure, the FHFF has robustness against model parameter uncertainties. We demonstrate the performance of the FHFF in comparison with existing nonlinear filters, such as the fuzzy Kalman filter and the particle filter.

Maximum likelihood FIR filter for visual object tracking

Visual object trackers usually adopt filters, such as the Kalman filter (KF) and the particle filter (PF), in order to improve tracking accuracy by suppressing measurement noises. However, if the filters have infinite impulse response (IIR) structures, the visual trackers adopting them can exhibit degraded tracking performance when system models have parameter uncertainties or when the noise information is incorrect. To overcome this problem, in this paper, we propose a new finite impulse response (FIR) filter for visual object tracking (VOT). The proposed filter is derived by maximizing the likelihood function, and it is referred to as the maximum likelihood FIR filter (MLFIRF). We conducted extensive experiments to show that the MLFIRF provides superior and more reliable tracking results compared with the KF, PF, and H filter (HF) in VOT. HighlightsAn alternative visual object tracker, called the maximum likelihood FIR tracker (MLFIRT), is proposed.The MLFIRT has the special robustness against model parameter uncertainties and incorrect noise information.Experimental results using the MLFIRT and the conventional visual trackers are presented.The experiments compare the MLFIRT with the Kalman tracker (KT) and the H-infinity tracker (HT).The MLFIRT provides superior and more reliable tracking results than the KT and the HT.

Particle filtering approach to membership function adjustment in fuzzy logic systems

The fuzzy logic system has been a popular tool for modeling nonlinear systems in recent years. In the fuzzy logic system, the shape of the membership function has a significant effect on the modeling accuracy. Thus, membership function adjustment methods have been studied and developed. However, in highly nonlinear systems, the existing membership function adjustment method based on the extended Kalman filter (EKF) may exhibit poor performance due to the linearization error. In this paper, to overcome the drawback of the EKF-based membership function adjustment (EKFMFA), we propose a new membership function adjustment method based on the particle filter (PF). The proposed PF-based membership function adjustment (PFMFA) does not suffer from performance degradation due to the linearization error. We demonstrate that the PFMFA outperforms the EKFMFA through the simulation of a fuzzy cruise control system.

A novel license plate character segmentation method for different types of vehicle license plates

License plate character segmentation (LPCS) is a very important part of vehicle license plate recognition (LPR) system. The accuracy of LPR system widely depends on two parts; namely license plate detection (LPD) and LPCS. Different country has different types and shapes of LPs are available. Based on character position on LP, we can find two types of LPs over the world, single row (SR) and double rows (DR) LP. Most of the LPCS methods are generally used for SRLP. This paper proposed a novel LPCS method for SR and DR types of LPs. Experimental results shows the real-time effectiveness of our proposed method. The accuracy of our proposed LPCS method is 99.05% and the average computational time is 27ms which is higher than other existing methods.

Detection and Recognition of Illegally Parked Vehicles Based on an Adaptive Gaussian Mixture Model and a Seed Fill Algorithm

In this paper, we present an algorithm for the detection of illegally parked vehicles based on a combination of some image processing algorithms. A digital camera is fixed in the illegal parking region to capture the video frames. An adaptive Gaussian mixture model (GMM) is used for background subtraction in a complex environment to identify the regions of moving objects in our test video. Stationary objects are detected by using the pixel-level features in time sequences. A stationary vehicle is detected by using the local features of the object, and thus, information about illegally parked vehicles is successfully obtained. An automatic alarm system can be utilized according to the different regulations of different illegal parking regions. The results of this study obtained using a test video sequence of a real-time traffic scene show that the proposed method is effective.

Modeling and Implementing Two-Stage AdaBoost for Real-Time Vehicle License Plate Detection

License plate (LP) detection is the most imperative part of the automatic LP recognition system. In previous years, different methods, techniques, and algorithms have been developed for LP detection (LPD) systems. This paper proposes to automatical detection of car LPs via image processing techniques based on classifier or machine learning algorithms. In this paper, we propose a real-time and robust method for LPD systems using the two-stage adaptive boosting (AdaBoost) algorithm combined with different image preprocessing techniques. Haar-like features are used to compute and select features from LP images. The AdaBoost algorithm is used to classify parts of an image within a search window by a trained strong classifier as either LP or non-LP. Adaptive thresholding is used for the image preprocessing method applied to those images that are of insufficient quality for LPD. This method is of a faster speed and higher accuracy than most of the existing methods used in LPD. Experimental results demonstrate that the average LPD rate is 98.38% and the computational time is approximately 49 ms.

Peak-to-peak exponential direct learning of continuous-time recurrent neural network models: a matrix inequality approach

The purpose of this paper is to propose a new peak-to-peak exponential direct learning law (P2PEDLL) for continuous-time dynamic neural network models with disturbance. Dynamic neural network models trained by the proposed P2PEDLL based on matrix inequality formulation are exponentially stable, with a guaranteed exponential peak-to-peak norm performance. The proposed P2PEDLL can be determined by solving two matrix inequalities with a fixed parameter, which can be efficiently checked using existing standard numerical algorithms. We use a numerical example to demonstrate the validity of the proposed direct learning law.

Performance comparison of stepwise parallel and serial cell search in WCDMA

For three-step cell search in WCDMA system, a stepwise serial scheme is conventionally employed, where each step operates in serial. In order to reduce cell search time, a stepwise parallel cell search scheme can be considered, where each step works in pipelined operation. However, in this parallel scheme where the processing time in every step is equal, excessive accumulations are caused in step (1) and step (3), because the code period for step (2) is much longer than those for the other steps and the effect of accumulations becomes saturated with the number of accumulations. This reduces the benefit of the parallel scheme. In this paper, the performance of parallel cell search is analyzed and compared with that of serial cell search over Rayleigh fading channels. The results are presented for varying parameters. Finally, it is shown that the performance of parallel cell search can be improved by adjusting appropriately the processing time in each step.