Deep learning-based hand gesture recognition for collaborative robots

Intention Prediction and Human Health Condition Detection in Reaching Tasks with Machine Learning Techniques

“…Tests were performed on the complete dataset of all the subjects, and compared with the results provided by the analysis of the subsets of healthy and pathological subjects only. For the comparison, the prediction accuracy of the LDA algorithm was calculated as depicted by Nuzzi et al [ 34 , 35 ]; for the assessment of the RF accuracy, the out-of-bag (OOB) method was adopted. The results were averaged on 200 consecutive tests.…”

“…The additional output of the left (L) and right (R) distinction was also evaluated in four tests, increasing to six the prediction classes for this kind of tests. Tests were performed on the complete dataset of all the subjects and the prediction accuracy of both the LDA and RF algorithms was evaluated according to Nuzzi et al [34,35]. For the assessment of the RF algorithm accuracy, the OOB approach was also used.…”

Efficient inception V2 based deep convolutional neural network for real‐time hand action recognition

“…Nuzzi et al [13] proposed a Faster R-CNN (object detection) and the custom prediction function (CPF) to estimate the confidence score of predictions. The confidence score of >90% is retained and rest is discarded.…”

“…Here too, the author did not report on the computational time. Nuzzi et al [2] used a Faster R-CNN to detect the position of the hands in the RGB images and average precision of 91.94% with the computation time of 230 ms. Nuzzi et al [13] proposed Faster R-CNN withCPF (Faster R-CNN + CPF) to recognise the hands with an average precision of 95.51% and the prediction time of 130 ms ± 10 ms. Kopuklu et al [14] achieved the state-of-the-art classification accuracy of 94.04%. Pisharady et al [16] reported a recognition rate of 94.36% with the prediction time of 2.65 s. Ji et al [17] proposed a classifier, which resulted in the recognition accuracy of 97.72%.…”

Inclusive Human Intention Prediction with Wearable Sensors: Machine Learning Techniques for the Reaching Task Use Case

“…x LDA and RF algorithms' prediction accuracy was computed and compared with respect to data sample dimension, number of considered features, and OW type. Accuracy was computed according to [21,22] and, in the case of RF accuracy, an out-of-bag (OOB) approach was also used. Table 2 depicts all the obtained results, averaged over 200 consecutive tests.…”