Search

Brillouin optical time domain analyzer (BOTDA) assisted by optimized support vector machine (SVM) algorithm for accurate temperature extraction is presented and experimentally demonstrated. Three typical intelligent optimization algorithms, particle swarm optimization algorithm, genetic algorithm and firefly algorithm are explored to optimize the SVM parameters. The performances of optimized SVM algorithms for temperature extraction are investigated in both simulation and experiment under different conditions for Brillouin gain spectrum collection, resulting in the significant enhancement of sensing accuracy. In particular, the extraction accuracy (i.e., smaller root mean square error value) of temperature information is improved about 4 °C compared with the conventional SVM when the signal-to-noise ratio (SNR) as low as 2.5 dB and 40-ns pump pulse width are adopted in the experiment. In addition to the enhanced accuracy with good robustness, the optimized algorithms have faster processing speed than the curve fitting method, over 20-times improvement. That makes the optimized algorithms become a very promising candidate for high performance BOTDA sensors in the future.

In this paper, harmonic concise atoms for the mean scatterer spacing (MSS) estimation was proposed to distinguish the normal and the ablated liver tissues. First, the second harmonic RF echo signals are separated with the Butterworth bandpass filter from the ultrasound RF echo signals, and a series of Gabor atoms are found to represent the local components with the concentrated energy. Then, the concise atoms are selected from the found Gabor atoms to optimal match the coherent components in the harmonic signals. Finally, the locations of the adjacent concise atoms are used to estimate the MSSs. In the microwave ablation experiments, the fundamental and second-harmonic RF echo signals were collected from the regions of interest in the normal untreated porcine tissues and thermally coagulated tissues ablated for 300 s, respectively. The MSSs, as well as the means and standard deviations of the MSSs were calculated from the collected signals for further comparation. The experiment results showed that the harmonic-based MSSs had large difference between the normal and the ablated liver tissues. The decrement in the harmonic-based mean MSS, and standard deviation in the ten tissues were 56.33 %, and 8.33 % greater than the fundamental-based results. It indicated that the harmonic-based MSS results with the concise atoms could improve the MSS estimation accuracy to distinguish the normal and the thermal coagulated liver tissues induced by the microwave ablation.

Infiltration growth to surrounding normal tissues is a major feature of malignant lesions, which makes benign and malignant breast lesions have significant differences in the margins and surrounding tissues. Entropy is a measurement for the uncertainty in a random variable, and has been used to quantitatively analyze the margins and surrounding tissues of benign and malignant breast lesions. In the present study, the entropy based on adaptively decomposed ultrasound radio frequency (RF) data is proposed for the characterization of breast lesions tissues. Down-sampling and dilation are used to preprocess raw ultrasonic RF signals, and then the fast multivariate empirical mode decomposition (FMEMD) is used to adaptively decompose the preprocessed data into a series of intrinsic mode functions (IMFs). The ring regions of interests (ROIs) that are the areas surrounding the tumors in all individual IMFs are determined to calculate entropies, respectively. The assessment is performed on Open Access Series of Breast Ultrasonic Data (OASBUD) with the RF data of 24 benign (breast imaging reporting and data system (BI-RADS) category 3) and 24 malignant (BI-RADS category 5) breast lesions. The results demonstrate that the decomposed entropies based on IMF3 and IMF4 have the superior performance in distinguishing benign and malignant breast lesions.

In ultrasound contrast imaging, the second harmonic (SH) of contrast agents has a higher resolution than the subharmonic and a higher energy than the ultraharmonic, which is helpful for the visualization of blood flow. However, the contrast-agent SH directly separated from the radio frequency signals may contain SH components generated by tissues. In this study, total least-squares (TLS) adaptive filtering is proposed to improve the second harmonic imaging (SHI) of ultrasound contrast agents. First, the pulse inversion based SH signals before and after injecting contrast agents are the reference signal and input of the TLS model, respectively. Then, the optimal model coefficients are determined by minimizing the perturbations in the input and output signals of the model. Finally, the contrast-agent SH is finally obtained by subtracting the optimal output from the input. In the simulation experiments, simulations before and after injecting contrast agents are used to evaluate the performance of the proposed method. Four random perturbations with different amplitude are added to the simulations after injecting contrast agents, respectively. The TLS-based contrast-to-tissue ratio (CTR) is the best when the tissues are static. For the perturbations of 0.5, 1.0 and 1.5 mm, the TLS-based CTRs are 13.1, 12.3 and 10.1 dB higher than the method combining pulse-inversion with wavelet transform imaging, respectively. In conclusion, the proposed TLS method suppresses more tissue components, and thus provides the improved performance in contrast-agent SHI.

Pulse wave velocity (PWV) is currently the gold standard for quantitatively measuring arterial stiffness, which is of great significance for the early diagnosis and effective prevention of cardio-cerebral vascular diseases. With the development of the ultrafast ultrasound imaging based on the plane wave technique, the ultrasonic transit time-based method with straightforward principle is preferable for the detection of the local PWV of carotid artery. However, the filtering parameter settings in this method are difficult to be flexibly controlled, and the results of pulse wave transit time estimations are not stable. Aiming at these problems, an adaptive local PWV estimation based on the particle swarm optimization algorithm is proposed. First, An improved particle swarm optimization (PSO) with a fitness function evaluated in terms of the Savitzky-Golay (SG) filter length, SG polynomial degree, tracking position and tracking points parameters in regional upstroke tracking (RUT) approach is used to find the most optimum parameter settings for PWV estimation. To quantitatively evaluate the proposed adaptive local PWV estimation method (PSO_RUT), an ultrasound simulation model for three age groups is used. Results show that the mean PWV deviations from theoretical values estimated by the proposed PSO_RUT method are just within 3.8%, and lower than that estimated by RUT method. Therefore, the PSO_RUT approach proposed in this study can provide effective means for the improvement on the accuracy of ultrasound-based carotid PWV estimation.

The rapid development of ultrafast ultrasound imaging based on the unfocused transmission of plane-wave has led to the widespread attention of clutter filtering technology, since the discrimination between tissue and blood motion is critical for non-contrast ultrafast Doppler imaging of blood flow. The increasingly used clutter filtering method, i.e., the singular value decomposition (SVD) is essentially a black-box technique, and the direct rejection of any singular vector corresponding to the first singular values will result in the potential loss of blood flow or residue of tissue activities. To cater for a better filtering performance, an adaptive clutter rejection method based on the fast multivariate empirical mode decomposition (FMEMD) is proposed, and compared to the high-pass filter (HPF) and SVD in the task of performing blood flow imaging and velocity profile estimation on signals collected from carotid arteries of 10 healthy human subjects. The results demonstrate the superiority of the proposed method over the state-of-the-art techniques, especially for discriminating between blood flow and tissue signals near the vessel walls.

In this paper, a new TbDyFe thin-film coated Surface Acoustic Wave (SAW) sensor is proposed for sensing current. A 150 MHz SAW device with delay line pattern is fabricated photolithographically on 128° YX LiNbO 3 substrate. Then the TbDyFe thin-film is deposited into the propagation path of surface acoustic wave by magnetron sputtering to build the sensing device. The reference sensing device with TbDyFe thin-film and the naked device are collected in a dual-channel differential oscillator so as to build the current sensor. When the magnetic field excited by the current is applied, the magnetostrictive effect produced by the TbDyFe thin-film causes the change of SAW propagation speed. And then the change of the corresponding oscillator frequency is used to characterize the current. Taking advantage of the TbDyFe thin-film such as high magnetostrictive coefficient and low coercive force, high sensitivity of 16.8 kHz/A is achieved in the experiments.

Median ensemble empirical mode decomposition (MEEMD) represents a remarkable improvement based on the ensemble empirical mode decomposition (EEMD) method for alleviating the mode splitting and mode mixing problem. However, the single use of the median operator generates tough problems including the higher reconstruction error, and the presence of burr in decomposition products. Aiming at addressing these problems while catering to a better time-frequency representation of the industrial oscillations, a median complementary EEMD (MCEEMD) method is proposed in this paper. In this work, the median operator and the mean operator are skillfully combined during the ensemble process. Through the study on simulation and typical industrial oscillation case, the effectiveness of MCEEMD is verified compared with existing methods, including EEMD, CEEMD and MEEMD.

P-type copper-doped zinc telluride (ZnTe:Cu) is a good candidate as a back contact of cadmium telluride (CdTe) solar cell. The deposition rate, transmittance and resistivity of ZnTe:Cu films deposited via target bias radio frequency (r.f.) sputtering was studied. The target bias voltage considerably influenced ZnTe:Cu film resistivity. In the meantime we find that post-deposition heat treatment (PDHT) significantly reduces the electrical resistivity of the ZnTe:Cu films, which is due to increases in both carrier concentration and mobility. It is inspiring for us to further improve the conductivity of ZnTe:Cu by applying the r.f. coupled d.c. sputtering and PDHT.

This paper presents an enhanced partially-parallel low-density parity-check (LDPC) decoder based on Consultative Committee for Space Data Systems (CCSDS) standard for high speed near Earth applications. Architectural level optimization is exploited to reduce hardware complexity and increase decoding throughput. To enchance the parallelism of the partially-parallel decoder, high circulant weight submatrix of the parity check matrix (PCM) are partitioned into several identity matrices with corresponding cyclic-shifted value and processed separately. Furthermore, messages corresponding to adjacent rows of the cyclic-shifted identity matrix are stored in one memory entry, and low-complexity data aligners are designed between memories and processing units to avoid memory access conflicts. Hardware implementation of a (8176,7154) LDPC code on Xilinx Virtex-5 XC5VLX330 FPGA shows that the proposed decoder can achieve a decoding throughput of 1.02 Gb/s at 10 iterations.

In this paper, we demonstrate how a deep neural network (DNN) can be used to compensate for nonlinearities and distortion effects introduced by the latest technology of 5G transmitters. A linearization approach based on neural networks can successfully cope with the challenges introduced not only by the high nonlinearity, wide bandwidth, and high frequency but also with challenges due to inter-PA crosstalk and load modulation. The device-under-test used in this experiment, is a state-of-the-art 5G 4x4 active phased array (APA) operating in very highly nonlinear regions at 28 GHz with a 100 MHz wide 3GPPbasestationsignalandwithOTAmeasuredsignalsusedfor training. Using the proposed DNN based linearization technique, an improvement of 11 % in error vector magnitude (EVM) and 10 dB adjacent channel leakage ratio (ACLR) are achieved which demonstrates the promising potential of this technique for emerging broadband communication systems such as 5G/6G and low earth orbit (LEO) satellite networks

This paper discusses the training of deep neutralnetworks (DNNs) for electromagnetic problems. The main concernsinclude how to modify EM problems to take the advantageof the deep learning techniques and how to tailor conventionaldeep learning concepts with electromagnetic domain knowledge,which has been overlooked by most existing DNN based EMresearch. A 1 8 patch antenna array has been adopted as thetest vehicle for investigation, with the aim to use deep learningfor radiation pattern synthesis. It is analyzed via electromagneticsimulation first to collect sufficient training data sets containingdifferent combinations of excitation signals and correspondingradiation patterns. These data are then pre-processed and passedto DNNs for training to imitate the mapping between excitationsignals and radiation patterns. With careful feature selection andDNN architecture optimizations, two DNN models are obtainedeventually. One of them aims at forward radiation synthesisin any certain excitation condition, and the other seeks outbackward excitation signals needed for a given radiation pattern,and both achieved an accuracy over 80%. This paper mayprovide enlightenment and reference in applying deep learningto electromagnetic problems in terms of feature selection andarchitecture modification.

This paper proposes a deep neural network (DNN)based digital signal recovery (DSR) technique for low Earth orbit (LEO) satellite communications. Different from existing work, which only investigates impacts of the satellite-to-ground communication channel, this work focuses on handling the nonlinearity variations caused by input power level perturbations in the transmitter. The system is validated using a high gain radio frequency power amplifier(RF-PA) operating at 28.5 GHz, where perturbations are introduced by varying the power level of the input signal, from −39 dBm to −31 dBm, to the RFPA. Experimental results show that the DNN trained at an input power level of−35 dBm achieved an improvement of 7.52 dB in the adjacent channel leakage ratio (ACLR), and an improvement of 4.2% in error vector magnitude (EVM). Applying the DDN trained at −35 dBm to other cases demonstrates that a 1 dB power level perturbation only leads to ≈ 1 dB degradation of the ACLR and ≈ 1.6% degradation of the EVM, respectively, which indicates the potential of the proposed approach. This paper proposes a deep neural network (DNN)-based digital signal recovery (DSR) technique for low Earth orbit (LEO) satellite communications. Different from existing work, which only investigates impacts of the satellite-to-ground communication channel, this work focuses on handling the nonlinearity variations caused by input power level perturbations in the transmitter. The system is validated using a high gain radio frequency power amplifier (RF-PA) operating at 28.5 GHz, where perturbations are introduced by varying the power level of the input signal, from -39 dBm to -31 dBm, to the RF-PA. Experimental results show that the DNN trained at an input power level of -35 dBm achieved an improvement of 7.52 dB in the adjacent channel leakage ratio (ACLR), and an improvement of 4.2% in error vector magnitude (EVM). Applying the DDN trained at -35 dBm to other cases demonstrates that a 1 dB power level perturbation only leads to ≈ 1 dB degradation of the ACLR and ≈ 1.6% degradation of the EVM, respectively, which indicates the potential of the proposed approach.

Domain knowledge assisted training dataset generation for deep learning aided metasurface designs is investigated. By combining the domain knowledge of metasurface from the designer and the advanced machine learning (ML) technique, an efficient training dataset generation approach has been successfully achieved. Unlike most existing metasurface generative designs that allow for arbitrary target pattern generation, which results in a time-consuming or even nonconvergence model training process, the proposed method takes the full advantages of the prior knowledge from designer to reduce the target solution space greatly, leading to reduced design cycles and higher ex-plainabilities of the designs. The proposed ML model combines generative adversarial network (GAN) and variational autoen-coder (VAE) as an encoder to transfer the original data into the latent space, which greatly improves the design efficiency as demonstrated in the validation results.

During unbalanced grid condition, the electromagnetic torque of the doubly fed induction generator (DFIG)-based wind turbine will contain oscillation at twice of the grid frequency, which is harmful to the gearbox and shafting system of the turbine. Though various control strategies have been proposed to suppress the electromagnetic torque ripple, it is still a difficult issue to achieve satisfactory dynamic response and low complexity, simultaneously. To settle this problem, this paper puts forward a low-complexity model predictive direct rotor current control (LC-MPDRCC) strategy, with the control delay being compensated and the effect of the parameters changes being discussed in detail. Simulation studies verify the effectiveness and feasibility of the proposed method.

In this paper, a high-performance dB-linear variable gain amplifier (VGA) based on gain interpolation is introduced. The VGA core uses a resistor ladder as an attenuator to get different attenuation steps of the input signal. By smoothly moving the gain from one step to the next step, an accurate dB-linear gain characteristic is achieved. A low noise amplifier and a post amplifier are integrated to get lower noise and ensure high linearity, respectively. The chip is fabricated in a 180nm SiGe process and occupies a die area of 1.15mm2. The total gain range is 48dB and the measured gain error is less than 0.27dB. The whole VGA consumes 30mA current from a 5V supply and the bandwidth is 120MHz. The measured in-band input referred noise density is 1nV/√HZ and the measured OIP3 is 33dBm.

The primary goal of an assist-as-needed (AAN) controller is to maximize subjects’ active participation during motor training tasks while allowing moderate tracking errors to encourage human learning of a target movement. Impedance control is typically employed by AAN controllers to create a compliant force-field around the desired motion trajectory. To accommodate different individuals with varying motor abilities, most of the existing AAN controllers require extensive manual tuning of the control parameters, resulting in a tedious and time-consuming process. In this paper, we propose a reinforcement learning AAN controller that can autonomously reshape the force-field in real-time based on subjects’ training performances. The use of action-dependent heuristic dynamic programming enables a model-free implementation of the proposed controller. To experimentally validate the controller, a group of healthy individuals participated in a gait training session wherein they were asked to learn a modified gait pattern with the help of a powered ankle-foot orthosis. Results indicated the potential of the proposed control strategy for robot-assisted gait training.

this paper proposes one feature-level target fusion method solving the problem of perception processing based on multi-source heterogeneity of target information. By mining the correlation between target data, basic data, meteorological data and hydrological data, the multi-source heterogeneous data could be effectively used. According to the results of testing in complex scenes, the proposed method could solve the difficult problem of occlusion interleaving and recognition redundancy which is involved by traditional single-source homogeneity perception. The proposed method could identify targets more accurately and have a good fusion effect.

According to the complex and changeable charging environment of mobile energy storage charging vehicles, this paper proposes an intelligent flexible charging strategy based on queuing theory for the single control strategy of traditional mobile energy storage charging vehicles. This strategy takes the optimal charging time as the optimization goal and dynamically adjusts the output power of the charger. At the same time, flexible charging also involves DC / DC module parallel current sharing and other issues. In this paper, a DCDC converter is used to control the charge and discharge of the battery pack. A charge and discharge controller for the energy storage battery pack based on PI control is designed. The energy storage battery pack During the charging process, the current closed-loop control realizes constant current charging, and the voltage outer loop current inner loop double closed loop control realizes constant voltage charging; during the discharge of the energy storage battery pack, the double closed loop is used to stabilize the battery pack discharge voltage, which is established in the Matlab environment The simulation model of charge and discharge of energy storage battery pack was obtained, and the simulation results verified the correctness and feasibility of the battery pack structure and control strategy.

Monocular pipelines are convenient and cheap solutions for object distance estimation in 3D vision. Current methods for monocular object distance estimation either perform inaccurately or require heavy work. In this paper, we propose a network with R-CNN based structure to implement object detection and distance estimation simultaneously. To improve the estimation results, we add a shallow network to process camera extrinsic parameters and optimize feature processing structure. We train and validate our network on KITTI object dataset, and compare with other methods to show that our method is accurate and easy to train. To prove the generality of our method under other scenarios, we construct a dataset of surveillance scenes, and conduct similar experiments on this dataset.

In view of the different probability of residents of different floors in high-rise buildings taking the elevator during the peak period of a day, this paper proposes an optimization scheduling model based on Monte Carlo Algorithm, making the existing model more accurate. The automatic counting module (ACM) based on infrared sensor is used to measure and record the number of passengers on different floors. On the basis of a large amount of data obtained from the research, the probability of taking the elevator is calculated, and the cubic polynomial fitting curve is generated based on the least square method, which is applied to the implementation of Monte Carlo Algorithm, and an optimized elevator scheduling model is obtained. At the same time, the optimized model and the existing model are compared in different aspects to generate visual comparison result. In the hypothesis of this paper, under the condition that the number of floors is 20 and the number of households on each floor is 50, the existing model will have a relative error of 10%-20%, and the number of passengers transported by the elevator below 10 floors will be significantly reduced with a relative error of more than 10%. At last, from the elevator scheduling model, a “one-to-many” model can be abstracted. We find that the model optimization idea used in this problem has certain guiding significance in dealing with some other practical problems.

For implementing mass customization on cloud manufacturing (CMfg), Internet of Things (IoT)-enabled service optimal composition (ISOC) is a key technology used to effectively composite a manufacturing cloud service with selected IoT services to satisfy the users’ customized production requirements. The solving of the ISOC problem is done inefficiently in the context of the growing scale of IoT services and increasing sophistication of the ISOC execution path, being partly due to insufficient investigation of accumulated empirical knowledge (EK) of IoT services. In this paper, we propose an EK-oriented genetic algorithm (EK-GA) for the large-scale IoT service composition. First, by fully considering the distinctive features of IoT service and service domain features, the EK of IoT services are richly explored to divide the service space. Second, EK-oriented optimization strategies in the initial population, operators, and fitness function are presented to improve the local and global search abilities of the genetic algorithm for solving ISOC problems. Finally, the effectiveness of EK-GA for solving real-world ISOC problems in a private CMfg is verified through three types of experiments. By exploiting EK of IoT services for ISOC problems, this work makes novel contributions for mass customization on CMfg and enriches the practice of EK-oriented intelligence optimization.