—Handwritten text line segmentation on real-world data presents significant challenges that cannot be overcome by any single technique. Given the diversity of approaches and the...
Abstract. In this paper we give some pedagogical explanation of those situations where the use of redundant or almost-redundant information can improve the recognition rates of pat...
A multilayer neural network based on multi-valued neurons (MLMVN) is a new powerful tool for solving classification, recognition and prediction problems. This network has a number ...
In the `missing data' approach to improving the robustness of automatic speech recognition to added noise, an initial process identifies spectraltemporal regions which are do...
A novel face recognition approach is proposed, based on the use of compressed discriminative features and recurrent neural classifiers. Low-dimensional feature vectors are extract...