Human action recognition plays an essential job in robotics as the robots need to have an understanding towards the information given to it. For it to have knowledge of a given task it needs to identify the action and also have the knowledge of the task.This paper proposes techniques for human action recognition. Here feature extraction method is used to derive the characteristics from the video and these characteristics are used by probabilistic neural network to classify the action.To depict the unique activities performed in various perspectives, static features are proposed for addressing action recognition.By extricating the comprehensive characteristics from various temporal scales few features are displayed as focal points which speak to the global spatial-temporal distribution. WEIZMAN data set is used in the experiment. The framework shouldn't be retrained when circumstances are changed which infers that the trained database can be applied in a number of conditions, for example, such as background changes and the viewing angle. To overcome these problem this paper is proposed.
Volume 11 | 07-Special Issue
Pages: 593-599