A Gesture Recognition approach to classifying Allergic Rhinitis gestures using Wrist-worn Devices : a multidisciplinary case study
Xenofon Aggelidis, Andreas Bardoutsos, Sotiris Nikoletseas, Nikolaos G. Papadopoulos, Christoforos Raptopoulos, Pantelis Tzamalis · 2020
In this paper, we propose a multidisciplinary Gesture Recognition case study using a Machine Learning approach for the detection and classification of allergic rhinitis-related gestures. Allergic diseases and especially allergic rhinitis are among the most common diseases in the world, mostly underappreciated, causing considerable impairment of daily activities, including job, and school productivity. For this reason, close monitoring and early recognition of symptoms worsening are considered essential. We hypothesize that recognizing allergic rhinitis to patients by such an approach may be a useful tool for such purpose.In our study, for the first time, the most common allergic rhinitis gestures are identified, based on patients' description and specialists' experience. Our data is retrieved by a large pool of active allergic rhinitis patients attending three specialized outpatient clinics in Greece. Gestures are recorded with the help of a wristband Bluetooth device incorporating a 3-axis accelerometer and a 3-axis gyroscope. Feature engineering and several signal processing methods are then applied to the raw sensor data (which are treated as 6-dimensional signals), and valuable features are extracted related to the time and frequency domains.To improve the performance of the Machine Learning models, we utilize Principal Component Analysis (PCA), and we also use functions such as Grid Search and Randomized Search, in order to achieve higher recognition accuracy by hyperparameter optimization.With these features and steps of processing, we built a classifier that can uniquely identify 15 allergic rhinitis gestures with an accuracy of 93% in a challenging variety of moves in the patient's head (nose, eye, ear). It is worth noting that allergic rhinitis gestures are more subtle, varied and spontaneous than other moves that have been considered in the literature so far. To the best of our knowledge, this is the first time that a Machine Learning approach is successfully applied in such a challenging field like respiratory diseases.