Vassilis Syrris

A lattice-based neuro-computing methodology for real-time human action recognition

This work describes a computational approach for a typical machine-vision application, that of human action recognition from video streams. We present a method that has the following advantages: (a) no human intervention in pre-processing stages, (b) a reduced feature set, (c) modularity of the recognition system and (d) control of the models complexity in acceptable for real-time operation levels. Representation of each video frame and feature extraction procedure are formulated in the lattice theory context. The recognition system consists of two components: an ensemble of neural network predictors which correspond to the training video sequences and one classifier, based on the PREMONN approach, capable of deciding at each time instant which known video source has potentially generated a new sequence of frames. Extensive experimental study on three well known benchmarks validates the flexibility and robustness of the proposed approach.

Detection and identification of human actions using Predictive Modular Neural Networks

The aim of the present study is to validate a 2D kinematic model of human body in providing considerable features that they could be used for human actions classification. Human motion can be termed as a non-rigid, articulated motion, with body parts being piecewise rigid, held together by joints. The presented approach uses the fact that the human body has certain anthropometric proportion and uses the anatomical shape representation of the non-rigid and articulated human body contour. The body joints and the different body parts are detected with help of prior anatomical knowledge and extracted silhouette. The result of this kinematics based approach is a simple 2D human stick figure. Features are extracted from this 2D model and used to represent the human body. In the training phase, each training video is represented by a neural network, while in classification phase, the Predictive Modular Neural Network (PREMONN) [12] time series classification algorithm is applied to classify the human actions.

Statistical descriptors for human actions classification

The objective of this study is to investigate alternative ways for representing suitably, with the fewest possible assumptions, the information derived from video recordings. It proposes a set of statistical descriptors capable of summarizing all the available information from each video frame. A sequence of such features expresses the object motion implicitly without the need for object detection techniques and tedious pre-processing. A video application such as the human action recognition is then tackled as a time-series classification problem. Neural networks are used for the time-series learning; when they are simulated with a new human action video, their predictions constitute the input a typical classifier would require, in order for it to decide which model (from the known time-series) has possibly generated this video.

Classification through hierarchical clustering and dimensionality reduction

This work describes a two-mode clustering hierarchical model capable of dealing with high dimensional data spaces. The algorithm seeks a transformed subspace which can represent the initial data, simplify the problem and possibly lead to a better categorization level. We test the algorithm on two hard classification problems, the phoneme and the pedestrian recognition; both are typical classification problems from real-life applications. Finally, the model is compared with many other algorithms.

Application of Fuzzy Lattice Neurocomputing (FLN) for Environmental Data Estimation

This paper addresses the problem of wind speed prediction at a particular location in the urban area of Thessaloniki, Greece, based on a data set containing wind parameter values at two other different locations. The problem is formulated and processed, and clusters in the data domain are identified by means of the Fuzzy Lattice Neurocomputing (FLN) methodology which can be used on any data set with the structure of a lattice. The results of the specific application are presented comparatively to past work, involving neural networks, on the same data set, demonstrating both the performance and the accuracy of the method.

A hierarchical predictive scheme for incremental time-series classification

This paper deals with the issue of gradual classification of a multivariate sequence where the number of candidate time-series generators is significantly high. It proposes a prediction scheme that consists of two components: a hierarchical structure which organizes the time-series models and a decision maker tool that assigns and evolves a respective hierarchy of probabilities; the latter expresses the current beliefs as to what the best model is for every hierarchical level. Experimentation in the domain of video-based human action recognition exhibits the capacity of the proposed approach to achieve efficient knowledge representation and real-time performance.