IMUs are gaining significant importance in the field of hand gesture analysis, trajectory detection and kinematic functional study. An Inertial Measurement Unit (IMU) consists of tri-axial accelerometers and gyroscopes which can together be used for formation analysis. The paper presents a novel classification approach using a Deep Neural Network (DNN) for classifying hand gestures obtained from wearable IMU sensors. An optimization objective is set for the classifier in order to reduce correlation between the activities and fit the signal-set with best performance parameters. Training of the network is carried out by feed-forward computation of the input features followed by the back-propagation of errors. The predicted outputs are analyzed in the form of classification accuracies which are then compared to the conventional classification schemes of SVM and kNN. A 3-5% improvement in accuracies is observed in the case of DNN classification. Results are presented for the recorded accelerometer and gyroscope signals and the considered classification schemes.
Sign Language is used by the deaf community all over world. Internationally, various research groups are working towards the development of an electronic sign language translator to enhance the accessibility of a signer. By employing intelligent models and wearable devices such as inertial measurement units (IMUs), continuous signs leading to the formation of a complete sentence can be recognized effectively. The work presented here proposes a novel one-dimensional deep capsule network (CapsNet) architecture for continuous Indian Sign Language recognition by means of signals obtained from a custom designed wearable IMU system. The IMU records tri-axial acceleration and turn rate, and orientation of the sensor is evaluated using a complementary filter. All the signals are used in the proposed deep learning network for learning and recognition of the signed sentences. The performance of the proposed CapsNet architecture is assessed by altering dynamic routing between capsule layers. Performance of the model is compared to that of foundational convolutional neural networks (CNNs) in terms of accuracy, loss, false predictions and learnt activations. The proposed CapsNet yields improved accuracy values of 94% (for 3 routing) and 92.50% (for 5 routings) in comparison to CNNs which yield 87.99%. Improved learning of the architecture is also validated by spatial activations depicting excited units at the predictive layer. For the purpose of evaluating relative performance and competitive nature of models, a novel non-cooperative pick game is constructed. The game presents a pick-andpredict competition between CapsNet and CNN constrained to a single strategy adoption. Both models compete with each other in order to reach their best responses. Higher value of Nash equilibrium for CapsNet as compared to CNN indicates the suitability of the proposed approach.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.