Human action recognition (HAR) is one of most significant research topics, and it has attracted the concentration of many researchers. Automatic HAR system is applied in several fields like visual surveillance, data retrieval, healthcare, etc. Based on this inspiration, in this chapter, the authors propose a new HAR model that considers an image as input and analyses and exposes the action present in it. Under the analysis phase, they implement two different feature extraction methods with the help of rotation invariant Gabor filter and edge adaptive wavelet filter. For every action image, a new vector called as composite feature vector is formulated and then subjected to dimensionality reduction through principal component analysis (PCA). Finally, the authors employ the most popular supervised machine learning algorithm (i.e., support vector machine [SVM]) for classification. Simulation is done over two standard datasets; they are KTH and Weizmann, and the performance is measured through an accuracy metric.
TopI. Introduction
Human Action Recognition (HAR)has attained a great research interest in recent years due to its wide variety applicability in different scenarios like Visual Surveillance, Gesture Recognition, Human Robot Interactions (HRI), Human -computer Interactions (HCI), analysis of behavior content based data retrieval etc. HAR is a very significant and an inspiring topic of research. One of the most popular problems with HAR is that the similar action may perform in several ways, even by similar person. One more serious issue in the HAR is the similarity of poses those appears in the same manner from multiple viewpoints. Hence the major problem is to determine an appropriate action representation method which is more discriminative (thus one can discriminate different actions) as well as generalized (thus the same actions can detected, even performed by different types of moving people and acquired from any view point) (M. Keestra et al. (2015)).
In earlier, the HAR is accomplished with the help of external peripherals like Gyroscopes, Accelerometers and Magnetometers etc. However, there is need of a physical interaction between sensors and human beings those were doing action. Thus, the main problem rises at the restricted movements due to the presence of connected wires and complex device settings (M. Jain et al. (2013)). To solve these issues, the new HAR has come into picture which is based on video sensors. In video based HAR, the activities are supervised with the help of video cameras. The main aim of HAR is to make the system to identify action by modelling the action through some mathematical computational algorithms. Similar to Human Visual System (HVS) these mathematical algorithms must produce a label after analyzing the entire or partial action from the video. In the computer vision oriented research, the development is done in the direction of developing such type of computational algorithms such that we can gain a great understanding about the actions of human beings from digital videos and images.
Due to the great significance of HAR in several real world applications, it has gained a lot of research interest and so many researchers concentrated in this direction. It has been studied from past decades and it is still a very challenging issue in real time applications. One of the most popular problems with HAR is that the similar action may perform in several ways, even by similar person. Another important issue is that the same human pose appears quite different when observed from different viewpoints. Hence the major problem is to determine an appropriate action representation method which is more discriminative (thus one can discriminate different actions) as well as generalized (thus the same actions can detected, even performed by different types of moving people and acquired from any view point) (Manish Khare et al., (2017)).
This chapter proposes a new HAR method for recognizing human actions from videos. Towards the recognition of human action, we have modeled an automatic system that takes an image as input, analyses it through signal processing approaches and then exposes the action present in it. At this contribution, we have proposed a new feature extraction method by combining the features of two individual filters; they are Gabor filter and wavelet filter. For dimensionality reduction, Principal Component Analysis (PCA) is employed and for classification, the Support Vector Machine (SVM) is used.
The remaining chapter structure is organized as follows; the literature survey on HAR is explored in section II. Further the details of developed method for HAR are explored in section III. The particulars of experimental analysis are explored in section IV and the conclusions are issued in the last section V.