The paper presents an approach towards human gender recognition system. The Silhouettes from Center for Biometrics and Security Research (CASIA) gait database are segmented in order to identify major body points and to generate corresponding point-light display. The features such as two dimensional coordinates of major body points and joint angles are extracted from the point-light display. The features are classified using Hidden Markov Model (HMM) and Support Vector Machines (SVM). The study yields a recognition rate of 69.18% and 76.79% with 100 subject data using HMM and SVM respectively. There has been a significant improvement in recognition accuracy using joint angles as the features.
In this paper we are proposing a text clustering method based on a well-known Word Sense Disambiguation (WSD) algorithm, the Lesk algorithm, to classify textual data by doing highly accurate Word Sense Disambiguation. The clustering of text data is thus primarily based on the context or meaning of the words used for clustering. The Lesk algorithm is used to return the sense identifiers for the words used to classify the text files by looking up the senses of a word in a Knowledge-Base similar to the English WordNet (enriched with more informative columns or fields for each synset [synonym set] of the English WordNet database), so as to greatly increase the chances of contextual overlap, thereby resulting in high accuracy of proper sense or context identification of the words. The proposed scheme has been tested on a number of heterogeneous text document datasets. The clustering results and accuracies, obtained using the proposed scheme, have been compared with the results obtained using the K-means clustering algorithm on the Vector Space Models generated for all the heterogeneous textual datasets. Experimental results show that our algorithm performs much better than the Vector Space Model (VSM) and K-means based approach. The technique will thus help the users much better in searching for meaningful contextual information from a highly diversified collection of textual information, which is a key task of the information overload problem.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.