Comparing with X-ray, ultrasound imaging has the advantages of no radiation and easy operation, etc., which is expected to be an effective alternative in clinical assessment for the human spine. At present, freehand scanning is mainly used in ultrasound imaging field, and recently robot-assisted scanning has emerged as a promising way to gradually reduce human intervention. However, the automatic recognition and scanning, as well as the control strategy of robots that guarantee the tight coupling between a probe and human skin, are still challenges. Moreover, the current measurement method of the spine is usually conducted based on plain radiographs, which lacks spatial structure information of spine. In this article, a new radiation-free robotic ultrasound system for 3-D imaging and a measurement method of the spine based on 3-D imaging are presented. A fully convolutional network (FCN) named fuse-Unet based on RGB and depth images captured by an RGB-depth (RGB-D) sensor Kinect, is proposed to realize the automatic recognition of human spine area and scanning path pre-planning. Then a six-degree-of-freedom robotic arm plays the role of a doctor to complete the automatic scanning along the pre-planning path, during which a normal-vector-based method and two force sensors are used to ensure that the probe well fits the spine area. Finally, 3-D ultrasound reconstruction and visualization of the spine are realized, based on which cobb angles are calculated for assessing the morphological structure of the spine. Phantom and in vivo experiments are conducted for evaluating the performance of the proposed system. Experimental results validate the practicability and accuracy of the system, indicating its potential for clinical application.

In this paper, we present a novel method by using a sliding window filter (SWF) for real-time kinematic (RTK)-visual-inertial navigation. Unlike other recent works that retain only the states of visual keyframes to reduce computational complexity, we additionally retain the GNSS states (i.e., position, orientation, and velocity of the body and inertial biases at the time of capturing GNSS measurements) in the SWF to construct more appropriate constraints between measurements and states. In order to make the method run as a real-time system, especially when the SWF contains numerous GNSS states, we propose a parallel elimination strategy in a predefined elimination ordering, which can solve the Gauss-Newton problem and simultaneously obtain the covariance for ambiguity resolution (AR). We reveal when and how the system improves the AR performance. Moreover, we analyze the observability of the system under different conditions. We also conduct experiments in the real world and compare the results with the state-of-the-arts. Experimental results show that the proposed method is able to achieve a higher and stabler fixed rate in GNSS challenging environments, and has better positioning performance with or without measurements of a base station. We have decided to publish the code of our work for the community ¹ .

It is challenging to design an equalizer for the complex time-frequency doubly-selective channel. In this paper, we employ the deep unfolding approach to establish an equalizer for the underwater acoustic (UWA) orthogonal frequency division multiplexing (OFDM) system, namely UDNet. Each layer of UDNet is designed according to the classical minimum mean square error (MMSE) equalizer. Moreover, we consider the QPSK equalization as a four-classification task and adopt minimum Kullback-Leibler (KL) to achieve a smaller symbol error rate (SER) with the one-hot coding instead of the MMSE criterion. In addition, we introduce a sliding structure based on the banded approximation of the channel matrix to reduce the network size and aid UDNet to perform well for different-length signals without changing the network structure. Furthermore, we apply the measured at-sea doubly-selective UWA channel and offshore background noise to evaluate the proposed equalizer. Experimental results show that the proposed UDNet performs better with low computational complexity. Concretely, the SER of UDNet is nearly an order of magnitude lower than that of MMSE.

Channel estimation becomes challenging in near-field extremely large-scale multiple-input multiple-output (XL-MIMO) systems, since the channel sparsity in the angular domain is destroyed. To avoid designing a near-field sparse dictionary and the estimation errors caused by sparse transformation, in this letter, we formulate the wideband near-field channel estimation problem as an image super-resolution (SR) problem, and propose a model-driven iterative SR channel estimate network (MDISR-Net) based on the Bayesian principle. Specifically, each layer of MDISR-Net is composed of a convolutional neural network (CNN) followed by a gradient descent network, which are corresponding to the prior sub-problem and channel sub-problem. Stack of multiple layers enables channel estimation to be resolved iteratively by solving two sub-problems. Our proposed scheme does not require a sparse dictionary and hence avoid the estimation errors caused by sparse transformation. The numerical results demonstrate that MDISR-Net significantly improves channel estimation accuracy.