Informatica

Basic motion structures of crowd aggregation and crowd dispersion are defined and a novel method for identifying these crowd behaviours is proposed. Based on integral optical flow, background and foreground are separated and intensive motion region is obtained. Crowd motion is analysed at pixel-level statistically for each frame to obtain quantity of pixels moving toward or away from each position and their comprehensive motion at each position. Regional motion indicators are computed and regional motion maps are formed to describe motions at region-level. Crowd behaviours are identified by threshold segmentation of regional motion maps.

The Autoregressive model-based digital inverse filtering technique is applied in non-invasive detection of vocal fold paralysis. The vocal tract filter is modelled using variable order (up to 20) AR model which is adequate to individual characteristics of human vocal properties. This postulates the more accurate estimation of the glottal flow, disturbances of which are direct evidence of the vocal fold paralysis.

Fuzzy C-Means (FCM) algorithm is one of the commonly preferred fuzzy algorithms for image segmentation applications. Even though FCM algorithm is sufficiently accurate, it suffers from the computational complexity problem which prevents the usage of FCM in real-time applications. In this work, this convergence problem is tackled through the proposed Modified FCM (MFCM) algorithm. In this algorithm, several clusters among the input data are formed based on similarity measures and one representative data from each cluster is used for FCM algorithm. Hence, this methodology minimizes the convergence time period requirement of the conventional FCM algorithm to higher extent. This proposed approach is experimented on Magnetic Resonance (MR) brain tumor images. Experimental results suggest promising results for the MFCM algorithm in terms of the performance measures.

The paper defines the decomposition problem of a mixture of time series into homogeneous components. First part deals with a solution based on Bayesian approach in the case of independent observations, the other part is devoted to a solution of on-line decomposition for a time series consisting of weakly stationary components.

The problem of change point detection when the properties of the random process observed suddenly begin changing slowly is considered. The most probable time moments of changes are investigated. Random processes are described by autoregression equations. The situation is studied when slow changes in the properties of a random process take place according to the linear law. An example of solving the problem is presented, realized by computer.