Shengxin Xu

A Diffraction Measurement Model and Particle Filter Tracking Method for RSS-Based DFL

Device-free localization (DFL) based on received signal strength (RSS) measurements functions by measuring RSS variation due to the presence of the target. The accuracy of a certain localization method closely depends on the accuracy of the measurement model itself. Existing models have been found not accurate enough under certain circumstances as they cannot explain some phenomena observed in DFL practices. In light of this, we propose a new model to characterize the RSS variation, which invokes diffraction theory and regards the target as a cylinder instead of a point mass. It is observed that the proposed model agrees well with experimental measurements, particularly when the target crosses the link or is in the vicinity of the link. Since the proposed measurement model is highly nonlinear, a particle filter-based tracking method is used to generate the approximate Bayesian estimate of the target position. As a performance benchmark, we have also derived the posterior Cramér-Rao lower bound of DFL for a diffraction model. A field test has shown that the proposed diffraction model may improve the tracking accuracy at least by 45% in a single-target case and by 27% in a double-target case.

Enhancing indoor radio tomographic imaging based on interference link elimination

Radio tomographic imaging (RTI) is a promising technique to localize and track the target without wearing any electronic device. However, the performance of traditional shadowing-based RTI (SRTI) degrades in indoor environments due to the existence of interference links caused by multipath. The interference links can bring false spots in the imaging results of RTI and make the true spot drift, resulting in position estimation error of the target. In this paper, we propose an interference link canceling technique to improve the performance of RTI where temporal and spatial properties of shadowed links are jointly used to detect the interference links. Since the spatial detection relies on the prior knowledge of the position of the target, we use Kalman filter to provide the position estimation. Moreover, a mean-shift clustering method is adopted to obtain the initial position estimation of the target. The experimental results demonstrate that the proposed enhanced SRTI (ESRTI) method outperforms the existing methods in terms of both image quality and tracking accuracy. We employ both temporal and spatial properties of shadowed links to detect the interference links.We utilize the targets dynamic to obtain the position prediction required by spatial detection.We propose to estimate the initial position of the target based on mean-shift clustering.The performance of the proposed method is greatly improved compared to those of existing methods.

Bayesian Device-Free Localization and Tracking in a Binary RF Sensor Network

Received-signal-strength-based (RSS-based) device-free localization (DFL) is a promising technique since it is able to localize the person without attaching any electronic device. This technology requires measuring the RSS of all links in the network constituted by several radio frequency (RF) sensors. It is an energy-intensive task, especially when the RF sensors work in traditional work mode, in which the sensors directly send raw RSS measurements of all links to a base station (BS). The traditional work mode is unfavorable for the power constrained RF sensors because the amount of data delivery increases dramatically as the number of sensors grows. In this paper, we propose a binary work mode in which RF sensors send the link states instead of raw RSS measurements to the BS, which remarkably reduces the amount of data delivery. Moreover, we develop two localization methods for the binary work mode which corresponds to stationary and moving target, respectively. The first localization method is formulated based on grid-based maximum likelihood (GML), which is able to achieve global optimum with low online computational complexity. The second localization method, however, uses particle filter (PF) to track the target when constant snapshots of link stats are available. Real experiments in two different kinds of environments were conducted to evaluate the proposed methods. Experimental results show that the localization and tracking performance under the binary work mode is comparable to the those in traditional work mode while the energy efficiency improves considerably.

Improving target tracking by incorporating shadowing fading

To accurately localize the target, wireless positioning systems use range measurements from different anchors, which are deployed around the monitored region in advance. In most applications where the target such as a human body has a certain volume, the target will shadow the wireless links comprised of the anchors, making the received signal strength (RSS) of these links suffer great loss. Therefore, in this paper, we propose a hybrid approach which integrates shadowing fading and range measurements of the links to enhance the performance of target tracking. Due to the nonlinearity of the shadowing loss model and the range model, particle filtering (PF) maximizing the posterior distribution of the target state is employed to fusion the two kinds of measurements. Moreover, the posterior Cramer-Rao lower bound (PCRLB) is derived to explore the best possible performance of the hybrid approach. Simulation results show that the tracking performance is significantly improved when the shadowing fading is exploited.

A Shadowing Loss Compensation Method for Hybrid RSS-Based Indoor Localization

Received signal strength (RSS) based localization schemes can be categorized as device-based localization (DBL) and device-free localization (DFL) in terms of the target to be located. DBL transforms RSS measurements of anchor-tag links to ranges and then obtains the targets position through trilateration. DFL infers an attenuation image from the targets body shadowing and regards the position with maximum attenuation as estimation. Existing hybrid method combining these two methods is able to improve the localization performance. However, the impact of the targets body shadowing loss on the anchor-tag links is not taken into consideration, worsening the localization accuracy of the DBL and hybrid method. In this paper, a new method is proposed to compensate the body shadowing loss by the exponential model on the basis of DFL result and the targets orientation. Experimental results validate that the compensation is beneficial and the proposed localization method is more reliable. Localization accuracy of the modified DBL is improved by 20% compared with the conventional DBL and the proposed hybrid localization method is 10% better than the conventional hybrid method.

An Image Restoration Method Using Matrix Transform and Gaussian Mixture Model for Radio Tomographic Imaging

Radio Tomographic Imaging (RTI) is an attractive technique for imaging the nonmetallic targets within wireless sensor network. RTI has been used in many challenging environments and situations. Due to the accuracy of Radio Tomographic Imaging system model and the interference between the wireless signals of sensors, the image obtained from the RTI system is a degraded target image, which cannot offer sufficient details to distinguish different targets. In this paper, we treat the RTI system as an image degraded process, and we propose an estimation model based on mixture Gaussian distribution to derive the degradation function from the shadowing-based RTI model. Then we use this degradation function to recover an original image by a method called constrained least squares filtering. So far, many imaging models have been proposed for localization; however, they do not have a satisfied imaging accuracy. Simulated and experimental results show that the imaging accuracy of our proposed method is improved, and the proposed method can be used in the real-time circumstances.

Towards robust and efficient device-free localization using UWB sensor network

Abstract Device-free localization (DFL) is the method of using distributed wireless sensors to localize the target without carrying any devices. Existing DFL methods leverage the variation of narrowband received signal strength (NRSS) which is vulnerable to multipath fading, and thus results in considerable performance degradation in indoor environments. Moreover, the inefficient sensor deployment of traditional DFL involves huge human efforts, which is not suitable for emergency scenarios. In this paper, we utilize sensors transmitting ultra-wideband (UWB) signals to solve both problems. We proposed two RSS variation estimation methods based on the channel impulse response (CIR) measurements provided by UWB sensors, which turn out to be more robust to multipath than NRSS due to the fine multipath resolution of UWB signals. We also employ a higher operating frequency to enhance the shadowing loss for mitigating the multipath effect. Additionally, satisfactory sensor self-localization is achieved under the cooperative localization framework owing to the accurate ranging capability of UWB sensors. We conducted experiments in three different environments to explore the feasibility of our method. The results show that the proposed method gains much better localization performance and requires less human efforts than narrowband DFL.

Shadow fading assisted device-free localization for indoor environments

Device-free localization (DFL) localizes the target by employing the variation of received signal strength (RSS) due to the presence of the target which wears no device. The existing fingerprint-based DFL (FDFL) method simply compares the online RSS variations with the radio map and does not care about how RSS variation is caused. However, we have found that when some links are shadowed by the target, even in indoor environments, the RSS of the links will experience large attenuation. Hence, in this paper, we propose to leverage shadowed links to enhance the performance of FDFL method. Specifically, we first detect the shadowed links in the monitored region, where the detection takes both the RSS variations and fade levels of the links into consideration. Afterwards, we reformulate the fingerprint matching by reducing the search space into the shadow region for improving the localization accuracy. Moreover, we propose a geometrical localization (GL) method if shadow regions of the detected shadowed links intersect. The experimental results show that the performance of the proposed method has been improved significantly compared to the traditional FDFL method.

Locating the Node by Exploiting Shadowing Fading

Shadowing fading is generally considered to be detrimental to node localization based on received signal strength (RSS). However, in this paper, we propose a new localization method which exploits shadowing fading to localize the node with unknown position. As we know that when a link comprised of the unknown node and an anchor node is blocked by an obstruction, the RSS of the link will be greatly attenuated. The shadowing fading can thus provide direction information with respect to the position of the unknown node because the largest shadowing loss is observed when the obstruction is located on the link. Motivated by this fact, an obstruction moves according to the predefined trajectory between an array of anchor nodes and the unknown node, resulting in large shadowing loss of the links. To localize the node, we extract the shadowing loss induced by the obstruction and employ the shadowing loss to estimate the time instants when the links are crossed. The link crossing time estimation is subsequently utilized to localize the unknown node by a least square (LS) method. The experimental result shows that the localization error of proposed method is about 0.33m, which is almost improved by 87% compared with conventional RSS-based localization method.

Robust multi-target tracking in RF tomographic network

Radio tomographic imaging (RTI) is a promising technique which allows localizing and tracking targets carrying no electronic devices. It utilizes the attenuation of wireless links to generate images of the change in the propagation field. Objects that obstruct the wireless signals in the field will lead to bright blobs in RTI image. For multi-target tracking, we employ clustering to obtain cluster observations to assign to targets. However, the blob corresponding to a target may be divided into several clusters in the process of clustering. The phenomenon is called over-clustering, i.e., there will be several cluster observations originated from the same target. Global nearest neighbor (GNN) which is popular in data association is optimal only under the assumption that only one cluster is originated from a target. However over-clustering will reduce the multi-target tracking performance of GNN. In this paper, the joint probabilistic data association (JPDA) method which is robust to over-clustering is proposed to improve the multi-target tracking performance when over-clustering is present. Real experiments are conducted in a monitored region surrounded by 20 RF sensors. When over-clustering is present, the experimental results show that the minimum tracking error of JPDA and GNN is 0.24m and 0.37m, respectively.