A modular, unsupervised neural network architecture that can be used for clustering and classification of complex data sets is presented. The adaptive fuzzy leader clustering (AFLC) architecture is a hybrid neural-fuzzy system that learns online in a stable and efficient manner. The system used a control structure similar to that found in the adaptive resonance theory (ART-1) network to identify the cluster centers initially. The initial classification of an input takes place in a two-stage process: a simple competitive stage and a distance metric comparison stage. The cluster prototypes are then incrementally updated by relocating the centroid position from fuzzy C-means (FCM) system equations for the centroids and the membership values. The operational characteristics of AFLC and the critical parameters involved in its operation are discussed. The AFLC algorithm is applied to the Anderson iris data and laser-luminescent finger image data. The AFLC algorithm successfully classifies features extracted from real data, discrete or continuous, indicating the potential strength of this new clustering algorithm in analyzing complex data sets.

No abstract

Electroencephalographic (EEG) signal texture content analysis has been proposed for early warning of an epileptic seizure'. This approach was evaluated by investigating the interrelationship between texture features and basic signal informational characteristics, such as Kolmogorov complexity and fractal dimension2. The comparison of several traditional techniques, including higher-order FIR digital filtering, chaos, autoregressive and FFT time-frequency analysis was also carried out on the same epileptic EEG recording3. The purpose of this study is to investigate whether wavelet transform can be used to further enhance the developed methods for prediction of epileptic seizures. The combined consideration of texture and entropy characteristics extracted from subsignals decomposed by wavelet transform are explored for that purpose. Yet, the novel neuro-fuzzy clustering algorithm'9 is performed on wavelet coefficients to segment given EEG recording into different stages prior to an actual seizure onset.The specific complexities of signals, such as derived from human brain EEGs, their nonstationary/nonlinear/nonanalytic feawres require not only the exploration of powerful, computational methods, based on different mathematical theories and paradigms, but also finding the ways for their integration and the development of new analysis tools specifically suited for these non-analytical highly complex situations. The very illdefined nature of the epileptic EEG signals associated with irregular transients involved calls for extensions of theory into new domains and of similar extensions of computer applications. Of particular concern here is the ability to detect those transients in advance of the epileptic seizure onset in real-time. The currently existing approaches and methodologies do not provide that capability. Despite the numerous positive results47 and investigations of EEG time series based on purely nonlinear dynamical characteristics (e.g. Lyapunov exponents and fractal dimension of the strange attractor), the limitation of these studies is that a certain brain functional state may often last only for a brief time period, whereas, in order to obtain reliable estimates for the mentioned characteristics, long-term time series are needed. Also, the algorithms involved are extremely sensitive to "noise", both electrical and physiologic in its nature.We have suggested an alternative signal visualization approach8 based on effective calculation of a paired samples statistical distribution, or second order histogram. Certain characteristics, reflecting "local texture" properties, are being derived and analyzed from the obtained distribution for particular signal segments. Unlike the mentioned nonlinear dynamics technique, ours does not require long-term segment consideration and specification of the embedded dimension of the phase space. The method proposed analyses local texture features. Various image texture features have been successfully used by many investigators employing second order statistics9"°. We have us...

No abstract

No abstract