Improved Detection of Human Respiration Using Data Fusion Basedon a Multistatic UWB Radar
This paper investigated the feasibility for improved detection of human respiration using data fusion based on a multistatic ultra-wideband (UWB) radar. UWB-radar-based respiration detection is an emerging technology that has great promise in practice. It can be applied to remotely sense the presence of a human target for through-wall surveillance, post-earthquake search and rescue, etc. In these applications, a human target's position and posture are not known a priori. Uncertainty of the two
... actors results in a body orientation issue of UWB radar, namely the human target's thorax is not always facing the radar. Thus, the radial component of the thorax motion due to respiration decreases and the respiratory motion response contained in UWB radar echoes is too weak to be detected. To cope with the issue, this paper used multisensory information provided by the multistatic UWB radar, which took the form of impulse radios and comprised one transmitting and four separated receiving antennas. An adaptive Kalman filtering algorithm was then designed to fuse the UWB echo data from all the receiving channels to detect the respiratory-motion response contained in those data. In the experiment, a volunteer's respiration was correctly detected when he curled upon a camp bed behind a brick wall. Under the same scenario, the volunteer's respiration was detected based on the radar's single transmitting-receiving channels without data fusion using conventional algorithm, such as adaptive line enhancer and single-channel Kalman filtering. Moreover, performance of the data fusion algorithm was experimentally investigated with different channel combinations and antenna deployments. The experimental results show that the body orientation issue for human respiration detection via UWB radar can be dealt well with the multistatic UWB radar and the Kalman-filter-based data fusion, which can be applied to improve performance of UWB radar in real applications. results in the experiments             . However, one experimental condition was too simplified to accord with the reality in most of the studies. For example, the volunteer subjects were positioned facing toward the radar in     11, 12] , and the test person in [5, 6] laid supine with their thorax directly towards the antennas. However, a human target's position and posture cannot be limited in advance in real scenarios. One study has shown that, among multiple postures, such as backing towards the antennas, the narrowband radar detected respiration was strongest when the test person was facing toward the radar  . The result is consistent with radar detection principles, according to which the thorax motion in the radar's radial direction mainly contributes to the detected respiration, and a human target's postural variation might result in a decrease of the radial component of the thorax motion. Additionally, the human breathing cross-section of UWB radar varies with the position . Thus, uncertainty of a human target's posture and position leads to a key issue affecting the human respiration detected by UWB radar, namely the detected respiration depends closely on body orientation with respect to the radar. Since a human target's body orientation cannot always be guaranteed to be facing toward the radar, the respiratory-motion response contained in UWB echoes is much weaker than expected. This will result in performance degradation of UWB radar when being applied in practice, especially for the applications of non-line-of-sight (NLOS) detection of human targets, such as through-wall surveillance or trapped victim search and rescue after an earthquake. According to the knowledge of the authors, only a few studies have considered this issue    . For example, a hidden Markov model was designed to infer the subject facing direction in  , and a setup comprising multiple UWB transceivers was proposed to solve the issue by choosing the channel with the highest signal quality . However, the two studies aimed at applying UWB radar in medicine, or, to be more specific, in sleep apnea monitoring. Their experiments were carried out in free space, in which detection of human respiration using UWB radar is relatively easy. Mainly for trapped victim detection in post-earthquake emergency rescue, the body orientation problem of UWB radar has been referred to, but not provided any solution  . Inspired by the multi-channel and multi-transceiving techniques in        19, 20] , this paper proposed a method to solve the body orientation issue based on a multistatic UWB radar. Compared with UWB monostatic radars, namely single transmitting and single receiving antenna that are collocated, various combinations of transmitting and receiving antennas in a multistatic UWB radar form a multisensory system that provides spatial diversity, redundancy information, multiplexing gain and so on        19, 20] . However, for human target detection, multistatic UWB radars are mainly used for moving target location and tracking      . There is no study on improved detection of human respiration using this type of radar. In the paper, the multistatic UWB radar took the form of impulse radios and comprised of one transmitting and four separated receiving antennas. Then, an adaptive Kalman filter was designed to fuse the UWB echo data from all of the receiving channels to detect the weak respiration contained in those data. A Kalman filter is one of the most widely used data fusion methods that has performed well in applications, like maneuvering target navigation, biomedical signal processing and so on       . Especially for biomedical signal processing, a Kalman filter has often been applied to extract or predict physiological signals, such as respiration and ECG    . Compared with these signals, human respiration detected by UWB radar in NLOS applications has a much lower signal to noise ratio. So, to investigate the feasibility of the method, two types of targets were used in the experiment. One was a volunteer that curled upon a camp bed behind a brick wall. The other was an artificial breathing object that imitated the thorax motion due to respiration and moved perpendicularly to the radar's radial direction. In the experiment, the radar detected the weak respiratory motion response for both the targets based on the Kalman-filter-based data fusion. Moreover, the radar's performance was experimentally investigated based on its single transmitting-receiving channels, and the data fusion algorithm's performance with different channel combinations and antenna deploys were experimentally investigated, too. The experimental results show that data fusion based on a multistatic UWB radar system is feasible to solve the weak respiration detection issue caused by the body orientation, and can be applied to improve the performance of UWB radar for human target sensing in real applications. The paper is organized as follows: Section 2 presents detailed information about the method, mainly describing the multistatic UWB radar, the Kalman-filter-based data fusion algorithm, and the experiment setup; Section 3 illustrates the results from the experiment. Finally, discussion and concluding remarks are given in Sections 4 and 5, respectively.