Explore chapters and articles related to this topic
Modeling Natural Behaviors for Human-Like Interaction with Robots
Published in Takayuki Kanda, Hiroshi Ishiguro, Human-Robot Interaction in Social Robotics, 2017
Takayuki Kanda, Hiroshi Ishiguro
The configuration of the developed robot system is shown in Figure 4.3. The interpretation processes in Figure 4.1 are achieved by the white-colored components. Meanwhile, the facilitation processes are achieved by the orange-colored components. Each process gets its input from either/both speech recognition or/and gesture recognition, and it controls Robovie’s behavior as its output. For speech recognition input, we used a speech recognition system [10] and attached a microphone to a human in order to avoid noise (we appreciate that there are many studies on noise-robust speech recognition that do not require microphones attached to people). The speech recognition system can recognize reference terms, object-color information, and error-alerting words in human utterances. On the other hand, gesture recognition is done with a motion-capturing system. In gesture recognition, the system handles two types of processes: pointing posture recognition and pointing motion detection. The following describes the details of implementing each process in Figure 4.1.
Virtual Environments
Published in Julie A. Jacko, The Human–Computer Interaction Handbook, 2012
Kay M. Stanney, Joseph V. Cohn
Tracking technology also allows for gesture recognition, in which human position and movement are tracked and interpreted to recognize semantically meaningful gestures (Turk 2002). Gestures can be used to specify and control objects of interest, direct navigation, manipulate the environment, and issue meaningful commands. Gesture tracking devices that are worn (e.g., gloves, bodysuits) are currently more advanced than passive techniques (e.g., computer vision), yet the latter hold much promise for the future, as they can provide more natural, noncontact, and less obtrusive solutions than those that must be worn; yet limitations need to be overcome in terms of accuracy, processing speed, and generality (Erol et al. 2007).
Applications of Computer Vision
Published in Manas Kamal Bhuyan, Computer Vision and Image Processing, 2019
The ability of computers to recognize hand gestures visually is essential for progress in human-computer interaction (HCI) [35]. Gesture recognition is a computer vision research problem, and it has many applications ranging from sign-language recognition to medical assistance to virtual reality, and many more. However, gesture recognition is extremely challenging not only because of its diverse contexts, multiple interpretations, and spatio-temporal variations but also because of the complex non-rigid properties of the hand [35].
An enhanced artificial neural network for hand gesture recognition using multi-modal features
Published in Computer Methods in Biomechanics and Biomedical Engineering: Imaging & Visualization, 2023
A human being’s gesture is an individual status, or series of states, of a particular human body part, most notably the hands and face. One of them is the frequent use of hand gestures to convey ideas instead of using spoken language (Zhang et al. 2020). For a computer to recognise human body language, hand gesture recognition is necessary (Al-Hammadi et al. 2020a). Nonverbal communication is done through hand gestures. It includes linguistic material that conveys a lot of information in sign language. Furthermore, human-computer interaction (HCI) systems heavily rely on it. That’s why there is a big market for tracking hand gestures automatically. Since the turn of the century, this area has drawn the interest of numerous researchers. The following factors (Kausar and Javed 2011) have increased the significance of automatic hand gesture recognition: (1) the increase of physically impaired people and (2) the widespread use of eyesight & voice-activated operating systems, especially those found in online gaming, video streaming interfaces, and interactive technology (Onan et al. 2016). It is critical to many HCI applications, including smart TV control, telesurgery, video gaming, and virtual reality (Kausar and Javed 2011; Onan et al. 2021a). The transformation of sign language is among the most significant uses of hand motion identification. To effectively express crucial human communication information and emotions, sign language hand gestures are constructed in a very complicated fashion (Al-Hammadi et al. 2020b).
Touchless Interfaces in the Operating Room: A Study in Gesture Preferences
Published in International Journal of Human–Computer Interaction, 2023
Naveen Madapana, Daniela Chanci, Glebys Gonzalez, Lingsong Zhang, Juan P. Wachs
Though the gesture recognition accuracies are relatively high, they are not enough for a real-time gesture recognition system. In other words, surgeons may get frustrated when performing gestures that are hard to recognize (low accuracies) as they need to repeat them multiple times before it is accurately recognized. In this regard, we developed a feedback and a fail-safe mechanism in order to make the system robust to the errors. This mechanism is facilitated by a dance pad (refer to Figure 3) that allows surgeons to visually navigate across the top predictions of the system and make a final decision. In more words, it further allows them to navigate through the top five predictions (refer to Figure 4), all existing commands if necessary, before selecting the final command using the dance pad.
A New Dataset and Neural Networks Models to Estimate the Joint Coordinates of Human Hand from Electromyographic Signals
Published in Cybernetics and Systems, 2022
S. Kirchhofer, T. Chateau, C. Bouzgarrou, J.-J Lemaire
In the field of prosthetic control, the literature focuses on gesture classification using sEMG signals (Li et al. 2019; Moura, Favieiro, and Balbinot 2016; Chen and Zhang 2019). One of the challenges is to extract relevant features from the signals to increase the classification accuracy (Phinyomark, Phukpattaranont, and Limsakul 2012; Tigra 2016). With the emergence of deep learning, this first step of signal processing tends to disappear as it is now part of classification algorithms. It was pointed out that a convolutional network can achieve the same result regardless if sEMG or features are sent in input (Côté-Allard et al. 2019). Actually, first layers adapt themselves to feed other layers with relevant features. Ding et al. (2018) proposed a convolutional neural network that performs a better classification without feature extraction. As the classic features are arbitrarily choose, it makes sense that they are not optimized for a specific task as gesture recognition. The feature computation eliminates some information in the name of a dimension reduction. In the same time, some convolutional layers have a good ability to learn the best way to extract relevant specification from the signal.