Discover human poses similarity and action recognition based on machine learning

Abstract

In the computer vision field, human action recognition depending on pose estimation recently made considerable progress, especially by using deep learning, which improves recognition performance. Therefore, it has been employed in various applications, including sports and physical activity follow-up. This paper presents a technique for recognizing the human posture in different images and matching their pose similarity. This aims to evaluate the viability of employing computer vision techniques to verify a person's body pose during exercise and determine whether the pose is executed properly. Exercise is one strategy we use to maintain our health throughout life. Gymnastics and yoga are two examples of this type of exercise. The proposed algorithm identifies human action by recognizing the body's key points. The OpenPose library has been used to detect 18 key points of the human body. The action classification task is performed using the support vector machine (SVM) algorithm. Then, the algorithm computes the similarity of the human pose by comparing a model image to a test image to determine the matching score. Evaluations show that our method can perform at a competitive or state-of-the-art performance on a number of body pose datasets.

Medication dispenser with touch screen and consumption authenticator for monitoring adherence to medication in the elderly

“…To perform the classification of actions, 1,000 images were collected for each class, using the COCO human poses dataset. For each image, the keypoints were When comparing the accuracy percentages for posture detection using the OpenPose and SVM algorithm of this system with that developed by [20], it can be seen that the accuracy of this dispenser is better, since 97.69% accuracy was obtained while in the other obtained 87%. This improvement is attributed to the dispenser's focus on detecting a specific posture, while in the other system, it is used to detect 4 different types of actions which require further analysis.…”

“…TensorFlow (Google's deep learning library) was used to create the action recognition classifier model and OpenPose [20] for human pose estimation using TensorFlow. OpenPose is based on a convolutional neural network with supervised learning in the Caffe deep learning framework that takes a color image as input and produces the 2D locations of anatomical landmarks for each person in the image as output [21].…”

“…At each stage, the PAF mapping with the input feature layer and keypoint heatmap is taken into account and known as 𝐿 𝑡 and 𝑆 𝑡 shown in ( 1) and ( 2). The acquired heatmap and PAFs of the prediction stages provide the position information and the direction vector of all connected articulation points [20]. At each subsequent stage, the predictions from the previous stage and the original image F-features are concatenated and used to produce refined predictions.…”