Search

Abstract Physical unclonable functions (PUFs) are of immense potential in authentication scenarios for Internet of Things (IoT) devices. For creditable and lightweight PUF applications, key attributes, including low power, high reconfigurability and large challenge-response pair (CRP) space, are desirable. Here, we report a ferroelectric field-effect transistor (FeFET)-based strong PUF with high reconfigurability and low power, which leverages the FeFET cycle-to-cycle variation throughout the workflow and introduces charge-domain in-memory computing. The proposed PUF cells are fabricated at 28 nm node, and the experimental measurements reveal high uniformity, uniqueness and repeatability. Remarkably, our PUF achieves near-ideal reconfigurability and ultra-low 1.89fJ per bit readout energy, significantly outperforming the state-of-the-art PUFs. Furthermore, we show that the PUF is robust against parameter variations and resilient to machine learning (ML) attacks. These performances highlight the great promise of the FeFET-based strong PUF as a feasible IoT security solution.

Real-time prediction of blood glucose levels (BGLs) in individuals with type 1 diabetes (T1D) presents considerable challenges. Accordingly, we present a personalized multitasking framework aimed to forecast blood glucose levels in patients. The patient data was initially categorized according to gender and age and subsequently utilized as input for a modified GRU network model, creating five prediction sub-models. The model hyperparameters were optimized and tuned after introducing the decay factor and incorporating the TCN network and attention mechanism into the GRU model. This step was undertaken to improve the capability of feature extraction. The Ohio T1DM clinical dataset was used to train and evaluate the performance of the proposed model. The metrics, including Root Mean Square Error (RMSE), Mean Absolute Error (MAE) and Clark Error Grid Analysis (EGA), were used to evaluate the performance. The results showed that the average RMSE and the MAE of the proposed model were 16.896 and 9.978 mg/dL, respectively, over the prediction horizon (PH) of 30 minutes. The average RMSE and the MAE were 28.881 and 19.347 mg/dL, respectively, over the PH of 60 min. The proposed model demonstrated excellent prediction accuracy. In addition, the EGA analysis showed that the proposed model accurately predicted 30-minute and 60-minute PH within zones A and B, demonstrating that the framework is clinically feasible. The proposed personalized multitask prediction model in this study offers robust assistance for clinical decision-making, playing a pivotal role in improving the outcomes of individuals with diabetes.

Coronary heart disease (CHD) is a dangerous condition that cannot be completely cured. Accurate detection of early coronary artery disease can assist physicians in treating patients. In this study, a prediction model called HY_OptGBM was proposed for predicting CHD by using the optimized LightGBM classifier. To optimize the LightGBM classifier, the hyperparameters of the LightGBM model were adjusted. In addition, its loss function was improved, and the model was trained using adjusted hyperparameters. In this study, the hyperparameters of the prediction model were optimized by applying the most advanced hyperparameter optimization framework (OPTUNA). The improved loss function is referred to as the focal loss (FL). In this study, a prediction model was evaluated by using CHD data from the Framingham Heart Institute. To evaluate the performance of the prediction model, various metrics, including precision, recall, F score, accuracy, MCC, sensitivity, specificity, and AUC, were used. The AUC value of the proposed model was 97.8%, which was better than that of other comparative models. The results demonstrate that the rate of early identification of CHD among the general population can be improved by utilizing the proposed method. This, in turn, could serve to mitigate the costs associated with the medical treatment of patients suffering from CHD.

This work reports a novel low-temperature poly-silicon thin-film-transistor-based pixel circuit for active-matrix neurostimulation. The pixel circuit consists of four transistors and one capacitor (4T1C) for programmable current-mode stimulation, which are designed for storing stimulation intensity information, simultaneously stimulating a large number of channels, and discharging stimulation electrodes. Due to the high mobility and low threshold voltages of the devices, the fabricated circuit occupies a pixel area of $200\times 200\,\,\mu \text{m}\,\,^{\mathrm{ 2}}$ , and delivers a stimulation current of $147 ~\mu \text{A}$ , sufficient to stimulate a neuron. The turn-on resistance of the fabricated transistor is below 6 $\text{k}\Omega $ , sufficient to be used as switches for bioelectronic applications. By employing a discharging switch transistor, the accumulated charges on the stimulation electrodes were released, and the electrode voltage was reduced to 0.08 V, thus mitigating corrosion. We demonstrated that two pixel circuits at different rows and columns can output stimuli simultaneously without a noticeable delay. This pixel circuit shows high potential to scale up as an active-matrix neurostimulation system with a high channel count.

In recent years, reinforcement learning (RL) has been widely used to solve multi-agent navigation tasks, and a high-fidelity level for the simulator is critical to narrow the gap between simulation and real-world tasks. However, high-fidelity simulators have high sampling costs and bottleneck the training model-free RL algorithms. Hence, we propose a Multi-Fidelity Simulator framework to train Multi-Agent Reinforcement Learning (MFS-MARL), reducing the total data cost with samples generated by a low-fidelity simulator. We apply the depth-first search to obtain local feasible policies on the low-fidelity simulator as expert policies to help the original reinforcement learning algorithm explore. We built a multi-vehicle simulator with variable fidelity levels to test the proposed method and compared it with the vanilla Soft Actor-Critic (SAC) and expert actor methods. The results show that our method can effectively obtain local feasible policies and can achieve a 23% cost reduction in multi-agent navigation tasks.

Feature extraction is an essential part of automatic speech recognition (ASR) to compress raw speech data and enhance features, where conventional implementation methods based on the digital domain have encountered energy consumption and processing speed bottlenecks. Thus, we propose a Mixed-Signal Processing (MSP) architecture to efficiently extract Mel-Frequency Cepstrum Coefficients (MFCC) features. We design MSP-MFCC to pre-process speech signals in the analog domain, which significantly reduces the cost of the analog-to-digital converter (ADC), as well as the computational complexity of the digital back-end. Moreover, MSP-MFCC eliminates the time-consuming Fourier transform in the conventional digital realization by improving processing flow. We fabricated the analog part based on 180nm CMOS mixed-signal technology, then measured the chip. The measured results show the energy consumption of MSP-MFCC is $0.72~\mu \text{J}$ /frame, and the processing speed is up to $45.79~\mu \text{s}$ /frame. MSP-MFCC achieves 95% energy saving and about $6.4\times $ speedup than state of the art. Further, by using the features extracted by MSP-MFCC, speech recognition simulation reaches the accuracy of 98.2%, which also keeps the leading performance to its current counterparts. The proposed MFCC extractor is competitive for integration in the ultra-low-power always-on wearable speech recognition applications.

Objective:To observe the effect of spine nerves muscle stimulation technique(SNMS)combined with scalp acupuncture on early stroke patients with shoulder-hand syndrome(SHS).Methods:A total of 44 patients with shoulder-hand syndrome after early stroke were divided into control group and observation group according to the random number table method, with 22 cases in each group. The control group was treated with scalp acupuncture(The scalp movement area was needled and the midline of the vertex, anterior and posterior oblique line of the parietal of the healthy side were located according to the brain functional area)combined with conventional physiotherapy(included passive joint activity, neural booster and neuromuscular electrical stimulation therapy), 2 hours and 45 minutes per day, 5 days per week. The observation group was treated with spine nerves muscle stimulation combined with scalp acupuncture, 2 hours and 30 minutes per day, 5 days per week. Fugl-Meyer(FMA)of the upper limb motor function, visual analogue scale(VAS)for pain and modified Barthel index(MBI)scores were recorded and compared before and after treatment.Results:There were no significant differences in FMA, VAS and MBI scores between the two groups before treatment(P>0.05).In the control group, the FMA score(27.41±6.82 vs 34.36±8.38, Pvs 2.23±1.23, Pvs 39.55±9.90, PPPPP

Objectives:To investigate the therapeutic effect of vocational rehabilitation training camp on patients with spinal cord injury (SCI) returning to society.Methods:A total of 34 patients with spinal cord injury who were hospitalized and received rehabilitation treatments in Xiangya Boai Rehabilitation Hospital from October 2017 to October 2019 were recruited and randomly divided into control group (16 cases) and observation group (18 cases) according to the random number table method. The control group received conventional rehabilitation treatment (including basic training, daily life activity training, traditional Chinese medicine therapy, etc.) for 6 hours a day, 5 days a week, for 12 weeks. The observation group added 2 weeks of vocational rehabilitation training camp on the basis of conventional rehabilitation treatment, 10 hours a day, 5 days a week, for 2 consecutive weeks, a total of 14 weeks. Patients in two groups were evaluated using general self-efficacy scale (GSES), community integration questionnaire (CIQ), and the lam assessment of stages of employment readiness (LASER) before and after treatment. The statuses of returning to work of patients were followed up by telephone six months after discharge.Results:There were no statistically significant differences between the two groups in GSES scores and CIQ scores before treatment (P> 0.05). There were statistically significant differences in GSES scores (20.89±4.86∶31.44±3.42), CIQ-family integration scores (4.39±1.24∶10.50±1.72), CIQ-social integration scores (2.78±0.94∶8.06±1.86), CIQ-production activity scores (0.50±0.62∶3.06±0.80) and LASER scores of the observation group before and after treatment (P< 0.05). There were statistically significant differences in GSES scores (22.06±5.53∶28.63±4.54), CIQ-family integration scores (4.69±1.14∶6.69±1.49), CIQ-social integration scores (3.75±0.93∶6.19±1.17), CIQ-production activity scores (0.75±0.68∶2.13±0.62) and LASER scores of the control group before and after treatment (P< 0.05). After treatment, there were statistically significant differences in GSES scores (31.44±3.42∶28.63±4.54), CIQ-family integration scores (10.50±1.72∶6.69±1.49), CIQ-social integration scores (8.06±1.86∶6.19±1.17), CIQ-production activity scores (3.06±0.80∶2.13±0.62), and LASER scores between the two groups (P< 0.05). Six months after discharge, the rate of returning to work of SCI patients in the observation group was significantly higher than that of the control group (P< 0.05).Conclusion:Vocational rehabilitation training camp can significantly improve the general self-efficacy, community integration, willingness to work and the rate of returning to work of patients with spinal cord injury.

Objective:To observe the clinical effect of DAVID spinal system core muscles training on patients with chronic lower back pain.Methods:Sixty patients with chronic low back pain were randomly divided into experimental group and control group, with 30 cases in each group. The control group was given interference current therapy, conventional waist core strength training (bridge movement, lying leg raises, and elbow plank) and sling exercise training (SET).The experimental group was trained in interference current therapy and DAVID spine system training (F110 lumbar/thoracic extension test trainer, F120 lumbar/thoracic spin test trainer, F130 lumbar/thoracic anterior flexion tester).A semi-quantitative Oswestry disability index (ODI) and quantitative DAVID system assessment were applied to analyze the result before and after treatment.Results:With six weeks' treatment, pain relief, the ability of personal care, lifting, walking, sitting, standing and traveling, quality of sleep and social life, as well as evaluation of waist range of motion (ROM) of both groups improved obviously compared with that before, and the parameters of the experimental group was much better than that of the control group (P

CPU has insufficient resources to satisfy the efficient computation of the convolution neural network (CNN), especially for embedded applications. Therefore, heterogeneous computing platforms are widely used to accelerate CNN tasks, such as GPU, FPGA, and ASIC. Among these, FPGA can accelerate the computation by mapping the algorithm to the parallel hardware instead of CPU, which cannot fully exploit the parallelism. By fully using the parallelism of the neural network’s structure, FPGA can reduce the computing costs and increase the computing speed. However, the development of FPGA requires great design skills. As a heterogeneous development platform, OpenCL has some advantages such as high abstraction level, short development cycle, and strong portability, which can make up for the lack of skilled designers. This paper uses Xilinx SDAccel to realize the parallel acceleration of CNN task, and it also proposes an optimizing strategy of single convolutional layer to accelerate CNN. Simulation results show that the calculation speed could be improved by adopting the proposed optimizing strategy. Compared with the baseline design, the strategy of single convolutional layer could increase the computing speed 14 times. Performance of the whole CNN task could be improved 2 times more than before, and the speed of image classification could attain more than 48 fps.

Insufficient power supply is a huge challenge for wireless body area network (WBAN). Body channel wireless power transfer (BC-WPT) is promising to realize multi-node high-efficiency power transmission for miniaturized WBAN nodes. However, the behavior of BC-WPT, especially in the multi-node scenario, is still lacking in research. In this paper, the inter-degeneration mechanism of a multi-node BC-WPT is investigated based on the intuitive analysis of the existing circuit model. Co-simulation in the Computer Simulation Technology (CST) and Cadence platform and experiments in a general indoor environment verify this mechanism. Three key factors, including the distance between the source and the harvester, frequency of the source, and area of the ground electrodes, are taken into consideration, resulting in 15 representative cases for simulation and experiments studies. Based on the simulation parameters, an empirical circuit model to accurately predict the received power of multiple harvesters is established, which fits well with the measurement results, and can further provide guidelines for designs and research on multi-node BC-WPT systems.

Three dimensional (3D) imaging technology has been widely used for many applications, such as human−computer interactions, making industrial measurements, and dealing with cultural relics. However, existing active methods often require both large apertures of projector and camera to maximize light throughput, resulting in a shallow working volume in which projector and camera are simultaneously in focus. In this paper, we propose a novel method to extend the working range of the structured light 3D imaging system based on the focal stack. Specifically in the case of large depth variation scenes, we first adopted the gray code method for local, 3D shape measurement with multiple focal distance settings. Then we extracted the texture map of each focus position into a focal stack to generate a global coarse depth map. Under the guidance of the global coarse depth map, the high-quality 3D shape measurement of the overall scene was obtained by local, 3D shape-measurement fusion. To validate the method, we developed a prototype system that can perform high-quality measurements in the depth range of 400 mm with a measurement error of 0.08%.

Energy is an important consideration in wireless sensor networks. In the current compression evaluations, traditional indices are still used, while energy efficiency is probably neglected. Moreover, various evaluation biases significantly affect the final results. All these factors lead to a subjective evaluation. In this paper, a new criterion is proposed and a series of tunable compression algorithms are reevaluated. The results show that the new criterion makes the evaluation more objective. Additionally it indicates the situations when compression is unnecessary. A new adaptive compression arbitration system is proposed based on the evaluation results, which improves the performance of compression algorithms.

Flexible electronic devices, such as the typical thin-film transistors, are widely adopted in the area of sensors, displayers, wearable equipment, and such large-area applications, for their features of bending and stretching; additionally, in some applications of lower-resolution data converters recently, where a trend appears that implementing more parts of system with flexible devices to realize the fully flexible system. Nevertheless, relatively fewer works on the computation parts with flexible electronic devices are reported, due to their poor carrier mobility, which blocks the way to realize the fully flexible systems with uniform manufacturing process. In this paper, a novel circuit architecture for image processing accelerator using Oxide Thin-film transistor (TFT), which could realize real-time image pre-processing and classification in the analog domain, is proposed, where the performance and fault-tolerance of image signal processing is exploited. All of the computation is done in the analog signal domain and no clock signal is needed. Therefore, certain weaknesses of flexible electronic devices, such as low carrier mobility, could be remedied dramatically. In this paper, Simulations based on Oxide TFT device model have demonstrated that the flexible computing parts could perform 5 × 5 Gaussian convolution operation at a speed of 3.3 MOPS/s with the energy efficiency of 1.83 TOPS/J, and realize image classification at a speed of 10 k fps, with the energy efficiency of 5.25 GOPS/J, which means that the potential applications to realize real-time computing parts of complex algorithms with flexible electronic devices, as well as the future fully flexible systems containing sensors, data converters, energy suppliers, and real-time signal processing modules, all with flexible devices.

Recently, a combination of non-invasive neuroimaging techniques and graph theoretical approaches has provided a unique opportunity for understanding the patterns of the structural and functional connectivity of the human brain (referred to as the human brain connectome). Currently, there is a very large amount of brain imaging data that have been collected, and there are very high requirements for the computational capabilities that are used in high-resolution connectome research. In this paper, we propose a hybrid CPU-GPU framework to accelerate the computation of the human brain connectome. We applied this framework to a publicly available resting-state functional MRI dataset from 197 participants. For each subject, we first computed Pearson's Correlation coefficient between any pairs of the time series of gray-matter voxels, and then we constructed unweighted undirected brain networks with 58 k nodes and a sparsity range from 0.02% to 0.17%. Next, graphic properties of the functional brain networks were quantified, analyzed and compared with those of 15 corresponding random networks. With our proposed accelerating framework, the above process for each network cost 80∼150 minutes, depending on the network sparsity. Further analyses revealed that high-resolution functional brain networks have efficient small-world properties, significant modular structure, a power law degree distribution and highly connected nodes in the medial frontal and parietal cortical regions. These results are largely compatible with previous human brain network studies. Taken together, our proposed framework can substantially enhance the applicability and efficacy of high-resolution (voxel-based) brain network analysis, and have the potential to accelerate the mapping of the human brain connectome in normal and disease states.