Search

Wireless sensing is an emerging technique which empowers wireless devices with additional sensing ability, that is, the ability to sense the target location, activity, gesture, vital signs, etc., in a device-free manner by analyzing the influence of the target on surrounding wireless signals. Benefiting from its excellent feature extraction and analysis capability, deep learning has emerged as a promising tool to realize wireless sensing. However, the labor intensive training efforts on collecting training samples or retraining a trained system limit its practical application. This article attempts to provide an integrated view on this field and discuss the feasibility of leveraging some new types of deep learning networks to accomplish wireless sensing tasks with less training efforts. Specifically, we first introduce the deep learning based wireless sensing framework, review the methods to construct radio images, and summarize different types of deep learning networks. Then, we focus on new types of deep learning networks which could reduce training efforts from two perspectives, that is, alleviating the efforts of retraining the system in a new scenario by learning a universal similarity evaluation ability utilizing a deep similarity evaluation networks, and reducing the efforts of collecting samples by generating virtual training samples using deep generative adversarial networks. Then, we elaborate existing challenges and forecast future research directions. Finally, we carry out experiments to evaluate the effectiveness of the introduced schemes using wireless gesture recognition as the case study. [ABSTRACT FROM AUTHOR]

Device-free wireless localization and activity recognition is an emerging technique, which could estimate the location and activity of a person without equipping him/her with any device. It deduces the state of a person by analyzing his/her influence on surrounding wireless signals. Therefore, how to characterize the influence of human behaviors is the key question. In this paper, we explore and exploit a radio image processing approach to better characterize the influence of human behaviors on Wi-Fi signals. Traditional methods deal with channel state information (CSI) measurements on each channel independently. However, CSI measurements on different channels are correlated, and thus lots of useful information involved with channel correlation may be lost. This motivates us to look on CSI measurements from multiple channels as a radio image and deal with it from the two-dimensional perspective. Specifically, we transform CSI measurements from multiple channels into a radio image, extract color and texture features from the radio image, adopt a deep learning network to learn optimized deep features from image features, and estimate the location and activity of a person using a machine learning approach. Benefits from the informative and discriminative deep image features and experimental results in two clutter laboratories confirm the excellent performance of the proposed system. [ABSTRACT FROM PUBLISHER]

Device-free wireless localization and activity recognition (DFLAR) is a new technique, which could estimate the location and activity of a target by analyzing its shadowing effect on surrounding wireless links. This technique neither requires the target to be equipped with any device nor involves privacy concerns, which makes it an attractive and promising technique for many emerging smart applications. The key question of DFLAR is how to characterize the influence of the target on wireless signals. Existing work generally utilizes statistical features extracted from wireless signals, such as mean and variance in the time domain and energy as well as entropy in the frequency domain, to characterize the influence of the target. However, a feature suitable for distinguishing some activities or gestures may perform poorly when it is used to recognize other activities or gestures. Therefore, one has to manually design handcraft features for a specific application. Inspired by its excellent performance in extracting universal and discriminative features, in this paper, we propose a deep learning approach for realizing DFLAR. Specifically, we design a sparse autoencoder network to automatically learn discriminative features from the wireless signals and merge the learned features into a softmax-regression-based machine learning framework to realize location, activity, and gesture recognition simultaneously. Extensive experiments performed in a clutter indoor laboratory and an apartment with eight wireless nodes demonstrate that the DFLAR system using the learned features could achieve 0.85 or higher accuracy, which is better than the systems utilizing traditional handcraft features. [ABSTRACT FROM PUBLISHER]