Human activity recognition using 2D skeleton data and supervised machine learning

Abstract

Vision‐based human activity recognition (HAR) finds its application in many fields such as video surveillance, robot navigation, telecare and ambient intelligence. Most of the latest researches in the field of automated HAR based on skeleton data use depth devices such as Kinect to obtain three‐dimensional (3D) skeleton information directly from the camera. Although these researches achieve high accuracy but are strictly device dependent and cannot be used for videos other than from specific cameras. Current work focuses on the use of only 2D skeletal data, extracted from videos obtained through any standard camera, for activity recognition. Appearance and motion features were extracted using 2D positions of human skeletal joints through OpenPose library. The approach was trained and tested on publically available datasets. Supervised machine learning was implemented for recognising four activity classes including sit, stand, walk and fall. Performance of five techniques including K‐nearest neighbours (KNNs), support vector machine, Naive Bayes, linear discriminant and feed‐forward back‐propagation neural network was compared to find the best classifier for the proposed method. All techniques performed well with best results obtained through the KNN classifier.

Human Motion Gesture Recognition Based on Computer Vision

“…In Figure 7, the PCKh results are shown in the test set using the MPII data set, and all results have a tolerance of 0.5. Compared with the methods in the literature [22,23] that do not consider the hierarchical Complexity structure, this method can show better results for most body parts estimation. e PCKh value of this method is better than the results of other methods for the average value of all parts and the value of the wrist and ankle parts.…”

“…Similarly, the model in this paper focuses on the hierarchical relationship of the human body structure, which can be better detected in key parts that are prone to large deformations. e accuracy of the comparison between the wrist part and the method in [22] can be explained this point. Hybrid pose machine [30] and pyramid stream network [31] are generally considered to be classic methods for training on LIP data sets.…”

“…ese two methods are also considered to the method proposed in this paper and is the closest comparable method. Although the literature [22] and the literature [23] can estimate the structure information of the human body more accurately, the model proposed in this paper can estimate the pose on challenging body parts such as elbows, wrists, knees, and ankles.…”

“…Alzahrani et al [21] comprehensively utilized the depth data and skeleton data provided by Kinect and effectively improved the real-time and robustness of gesture recognition through anthropometric knowledge and back propagation neural network. Ghazal et al [22] used the Kinect camera to collect human bone information and 3D data corresponding to joint points for action recognition. Franco et al [23] also used Kinect equipment to obtain depth information and designed a human posture recognition system specifically for sitting and standing postures.…”

Motion Direction Inconsistency‐Based Fight Detection for Multiview Surveillance Videos

“…Generally, the behavior analysis methods are categorized in two classes: the methods based on high-level human body structure and the methods based on low-level image information. The information used in high-level methods includes body point model, 2D model of human body [2], and 3D model of human body [3][4][5]. Moreover, in order to obtain more information, multiview videos or a 3D camera are required for this method.…”