Highly-Optimized Radar-Based Gesture Recognition System with Depthwise Expansion Module

Abstract

The increasing integration of technology in our daily lives demands the development of more convenient human–computer interaction (HCI) methods. Most of the current hand-based HCI strategies exhibit various limitations, e.g., sensibility to variable lighting conditions and limitations on the operating environment. Further, the deployment of such systems is often not performed in resource-constrained contexts. Inspired by the MobileNetV1 deep learning network, this paper presents a novel hand gesture recognition system based on frequency-modulated continuous wave (FMCW) radar, exhibiting a higher recognition accuracy in comparison to the state-of-the-art systems. First of all, the paper introduces a method to simplify radar preprocessing while preserving the main information of the performed gestures. Then, a deep neural classifier with the novel Depthwise Expansion Module based on the depthwise separable convolutions is presented. The introduced classifier is optimized and deployed on the Coral Edge TPU board. The system defines and adopts eight different hand gestures performed by five users, offering a classification accuracy of 98.13% while operating in a low-power and resource-constrained environment.

Universal Adversarial Attacks on the Raw Data From a Frequency Modulated Continuous Wave Radar

“…In previous approaches, radar data is usually preprocessed before the information is transferred to an ANN in the form of Range-Doppler-Maps (RDMs) or other representations. Various applications of radar sensors are conceivable, ranging from distance determination and classification of objects in road traffic [11], to gesture recognition in consumer electronics [12] and even breath and heartbeat frequency detection of humans [13]. There is also an increasing trend to move the previous data preprocessing steps to the ANN as well, [14] as this allows specialized AI accelerators to shorten response times and enable processing on the edge [15], [16].…”

“…The dataset is a subset of the one presented by M. Chmurski [12]. Each gesture consists of 10 frames with the information of 3 antennas recording with 64 CPF and 32 SPC.…”

MM-Wave Radar-Based Recognition of Multiple Hand Gestures Using Long Short-Term Memory (LSTM) Neural Network

“…However, machine learning algorithms used in radar signal processing are still a nascent field. Current literature reports a classification of up to 10 motion gestures with >90% accuracy [7,[10][11][12][13][14][15][16][17][18][19]. In [10], it is discussed that the accuracy of such algorithms may drop by up to 40% when the classification is executed on samples from a subject not included in the training set.…”

Stepped-Frequency Continuous-Wave Signal Processing Method for Human Detection Using Radars for Sensing Rooms through the Wall

“…The heartbeat is a more regular process in the short term; it includes regular changes in the volumes of ventricles and atria of the heart [13] and subsequent pulse waves mostly affecting the main arteries, resulting in a total displacement measured as 0.1 cm to 0.5 cm. The class of vital signs generally described as the body movement is rather wide since it includes the variety of muscle-driven motions ranging from deliberate ones, such as gestures [14], to natural tremors [15,16], characterized mostly as realization of a non-stationary random process [17,18]. In the signal-processing view, signals reflected from a non-moving human tend to cause interframe oscillations in the amplitude and phase of the value measured in the resolution cells where the target is continuously observed.…”