This study presents a framework that recognizes and imitates human upper-body motions in real time. The framework consists of two parts. In the first part, a transformation algorithm is applied to 3D human motion data captured by a Kinect. The data are then converted into the robot's joint angles by the algorithm. The human upper-body motions are successfully imitated by the NAO humanoid robot in real time. In the second part, the human action recognition algorithm is implemented for upper-body gestures. A human action dataset is also created for the upper-body movements. Each action is performed 10 times by twenty-four users. The collected joint angles are divided into six action classes. Extreme Learning Machines (ELMs) are used to classify the human actions. Additionally, the Feed-Forward Neural Networks (FNNs) and K-Nearest Neighbor (K-NN) classifiers are used for comparison. According to the comparative results, ELMs produce a good human action recognition performance.

Özetçe-Bu çalışmada, insan hareketlerinin tanınması ve insansı bir robot üzerinde gerçekleştirilmesine yönelik yeni bir algoritma geliştirilmiştir. İki kısımdan oluşan çalışmanın birinci kısmında, eş zamanlı bir taklit sistemi sunulmaktadır. Xbox 360 Kinect algılayıcıdan elde edilen üç boyutlu iskelet eklem pozisyonları, bir dönüşüm algoritması sayesinde robotun kol eklem açılarına dönüştürülmekte ve sonrasında bu açılar, NAO insansı robota aktarılmaktadır. Bu kısmın sonunda, insan üst vücut hareketleri, NAO robot tarafından eş zamanlı olarak başarılı bir şekilde taklit edilmiştir. Çalışmanın ikinci kısmında, insan hareketlerinin tanınması için bir algoritma oluşturulmuştur. İnsan hareketlerini sınıflandırmak için Aşırı Öğrenme Makinaları (AÖM) ve geriye yayılım algoritması ile eğitilen ileri beslemeli Yapay Sinir Ağları (YSA) kullanılmıştır. Karşılaştırmalı sonuçlara göre, AÖM ile daha iyi bir tanıma başarımı elde edilmiştir. Anahtar Kelimeler -İnsan hareketlerinin tanınması; NAO insansı robot; Xbox 360 Kinect.Abstract-In this study, a novel algorithm is developed to recognize human actions and reproduce human actions on a humanoid robot. The study consists of two parts. In the first part, the real time human imitation system is realized. The three dimensional skeleton joint positions obtained from Xbox 360 Kinect. These positions are transformed to joint angles of robot arms via a transformation algorithm and these angles are transferred to NAO robot. The human upper body movements are finally successfully imitated by NAO robot in real time. In the second part, the algorithm is generated for the recognition of human actions. Extreme Learning Machines (ELMs) and the Feed Forward Neural Networks (FNNs) with back propagation algorithm are used to classify actions. According to the comparative results, ELMs produce a better recognition performance.

Abstract:As the behavior of a chaotic Chua's circuit is nonstationary and inherently noisy, it is regarded as one of the most challenging applications. One of the fundamental problems in the prediction of the behavior of a chaotic Chua's circuit is to model the circuit with high accuracy. The current paper presents a novel method based on multiple extreme learning machine (ELM) models to learn the chaotic behavior of the four elements canonical Chua's circuit containing a memristor instead of a nonlinear resistor only by using the state variables as the input. In the proposed method four ELM models are used to estimate the state variables of the circuit. ELMs are first trained by using the data spoilt by noise obtained from MATLAB models of a memristor and Chua's circuit. A multistep-ahead prediction is then carried out by the trained ELMs in the autonomous mode. All attractors of the circuit are finally reconstructed by the outputs of the models. The results of the four ELMs are compared to those of multiple linear regressors (MLRs) and support vector machines (SVMs) in terms of scatter plots, power spectral density, training time, prediction time, and some statistical error measures. Extensive numerical simulations results show that the proposed system exhibits a highly accurate multistep iterated prediction consisting of 1104 steps of the chaotic circuit. Consequently, the proposed model can be considered a promising and powerful tool for modeling and predicting the behavior of Chua's circuit with excellent performance, reducing training time, testing time, and practically realization probability.

No abstract

No abstract