This paper is a first step towards a smart hand gesture recognition set up for Collaborative Robots using a Faster R-CNN Object Detector to find the accurate position of the hands in RGB images. In this work, a gesture is defined as a combination of two hands, where one is an anchor and the other codes the command for the robot. Other spatial requirements are used to improve the performances of the model and filter out the incorrect predictions made by the detector. As a first step, we used only four gestures.

The evaluation of the morphological characteristics of small resistance arteries in humans is challenging. The gold standard method is generally considered to be the measurement by wire or pressure micromyography of the media-to-lumen ratio of subcutaneous small vessels obtained by local biopsies. However, noninvasive techniques for the evaluation of retinal arterioles were recently proposed; in particular, two approaches, scanning laser Doppler flowmetry (SLDF) and adaptive optics, seem to provide useful information; both of them provide an estimation of the wall-to-lumen ratio (WLR) of retinal arterioles. Moreover, a noninvasive measurement of basal and total capillary density may be obtained by videomicroscopy/capillaroscopy. No direct comparison of these three noninvasive techniques in the same population was previously performed; in particular, adaptive optics was never validated against micromyography. In the current study, we enrolled 41 controls and patients: 12 normotensive lean controls, 12 essential hypertensive lean patients, nine normotensive obese patients and eight hypertensive obese patients undergoing elective surgery. All patients underwent a biopsy of subcutaneous fat during surgery. Subcutaneous small resistance artery structure was assessed by wire micromyography and the media-to-lumen ratio was calculated. WLR of retinal arterioles was obtained by SLDF and adaptive optics. Functional (basal) and structural (total) microvascular density was evaluated by capillaroscopy before and after venous congestion. Our data suggest that adaptive optics has a substantial advantage over SLDF in terms of evaluation of microvascular morphology, as WLR measured with adaptive optics is more closely correlated with the M/L of subcutaneous small arteries (r = 0.84, P < 0.001 vs. r = 0.52, P < 0.05, slopes of the relations: P < 0.01 adaptive optics vs. SLDF). In addition, the reproducibility of the evaluation of the WLR with adaptive optics is far better, as compared with SLDF, as intraobserver and interobserver variation coefficients are clearly smaller. This may be important in terms of clinical evaluation of microvascular morphology in a clinical setting, as micromyography has substantial limitations in its clinical application due to the local invasiveness of the procedure.

This paper presents a new method for the automated processing of surface electromyography (SEMG) signals, particularly suited for the detection of muscle activation timing. The method has an intermediate level of complexity between simpler (but less performing) and more complex (but in general slower) methods, and is successfully used in the development of biomedical devices for rehabilitation carried out by our group. The method proposed here is based on a statistical approach for threshold computation that is implemented without the need of maximum voluntary contraction or relaxed state, usually required to overcome the difficulty in obtaining the threshold value. The method is compared with 10 popular automated standard methods using different types of simulated signals that approximate the behavior of real SEMG signals. Both the number of activations detected and the onset time measured are analyzed. The algorithm is then applied to real SEMG signals acquired from healthy subjects. The results are finally compared with the literature values. The results show that the proposed algorithm is the best performing method when both the number of activations and the activation timing are considered. In real applications, the algorithm gives the results compatible with the well-agreed literature data.

The PMD Camboard Picoflexx Time-of-Flight (ToF) camera is evaluated against the Microsoft Kinect V2 to assess its performance in the context of markerless motion capture system for human body kinematics measurements. Various error sources such as the warm-up time, the depth distortion, the amplitude related error, the signal-to-noise ratio, and limitations such as their dependence on the illumination pattern and on the target distance, are studied and compared. The Picoflexx device is also compared to the Kinect V2 in relation to the quality of shape reconstructions, to assess its adequateness in modeling human body segments, and human body kinematics measurements. The final result of this paper is definitely useful to the research community, demonstrating that, even if the Picoflexx performs lower than the Kinect concerning the measurement performances, its behavior in estimating the volume of the body segments, the angles at the joints for human body kinematics, and the angle at the ankle in assisted walking applications is definitely satisfactory. These results are extremely significant to obtain accurate estimates of the parameters of the human body models in markerless motion capture applications, especially in laboratory-free environments, where compactness, lightweight, wireless connection, and low-power consumption are of outmost importance.

The HANDS dataset has been created for human-robot interaction research, and it is composed of spatially and temporally aligned RGB and Depth frames. It contains 12 static single-hand gestures performed with both the right-hand and the left-hand, and 3 static two-hands gestures for a total of 29 unique classes. Five actors (two females and three males) have been acquired performing the gestures, each of them adopting a different background and light conditions. For each actor, 150 RGB frames and their corresponding 150 Depth frames per gesture have been collected, for a total of 2400 RGB frames and 2400 Depth frames per actor.