Reducing the process variation is a significant concern for resistive random access memory(RRAM).Due to its ultrahigh integration density,RRAM arrays are prone to lithographic variation during the lithography process,...Reducing the process variation is a significant concern for resistive random access memory(RRAM).Due to its ultrahigh integration density,RRAM arrays are prone to lithographic variation during the lithography process,introducing electrical variation among different RRAM devices.In this work,an optical physical verification methodology for the RRAM array is developed,and the effects of different layout parameters on important electrical characteristics are systematically investigated.The results indicate that the RRAM devices can be categorized into three clusters according to their locations and lithography environments.The read resistance is more sensitive to the locations in the array(~30%)than SET/RESET voltage(<10%).The increase in the RRAM device length and the application of the optical proximity correction technique can help to reduce the variation to less than 10%,whereas it reduces RRAM read resistance by 4×,resulting in a higher power and area consumption.As such,we provide design guidelines to minimize the electrical variation of RRAM arrays due to the lithography process.展开更多
The instantaneous reversible soft logic upset induced by the electromagnetic interference（EMI） severely affects the performances and reliabilities of complementary metal–oxide–semiconductor（CMOS） inverters. This...The instantaneous reversible soft logic upset induced by the electromagnetic interference（EMI） severely affects the performances and reliabilities of complementary metal–oxide–semiconductor（CMOS） inverters. This kind of soft logic upset is investigated in theory and simulation. Physics-based analysis is performed, and the result shows that the upset is caused by the non-equilibrium carrier accumulation in channels, which can ultimately lead to an abnormal turn-on of specific metal–oxide–semiconductor field-effect transistor（MOSFET） in CMOS inverter. Then a soft logic upset simulation model is introduced. Using this model, analysis of upset characteristic reveals an increasing susceptibility under higher injection powers, which accords well with experimental results, and the influences of EMI frequency and device size are studied respectively using the same model. The research indicates that in a range from L waveband to C waveband, lower interference frequency and smaller device size are more likely to be affected by the soft logic upset.展开更多
Hummingbirds have a unique way of hover- ing. However, only a few published papers have gone into details of the corresponding three-dimensional vortex struc- tures and transient aerodynamic forces. In order to deepen...Hummingbirds have a unique way of hover- ing. However, only a few published papers have gone into details of the corresponding three-dimensional vortex struc- tures and transient aerodynamic forces. In order to deepen the understanding in these two realms, this article presents an integrated computational fluid dynamics study on the hovering aerodynamics of a rufous hummingbird. The original morphological and kinematic data came from a former researcher＇s experiments. We found that conical and sta- ble leading-edge vortices （LEVs） with spanwise flow inside their cores existed on the hovering hummingbird＇s wing surfaces. When the LEVs and other near-field vortices were all shed into the wake after stroke reversals, periodically shed bilateral vortex rings were formed. In addition, a strong downwash was present throughout the flapping cycle. Time histories of lift and drag were also obtained. Combining the three-dimensional flow field and time history of lift, we believe that high lift mechanisms （i.e., rotational circulation and wake capture） which take place at stroke reversals in insect flight was not evident here. For mean lift throughout a whole cycle, it is calculated to be 3.60 g （104.0 % of the weight support）. The downstroke and upstroke provide 64.2 % and 35.8 % of the weight support, respectively.展开更多
Epidermal sensing devices,which mimic functionalities and mechanical properties of natural skin,offer great potential for real-time health monitoring via continuous checking of vital signs.However,most existing skin-m...Epidermal sensing devices,which mimic functionalities and mechanical properties of natural skin,offer great potential for real-time health monitoring via continuous checking of vital signs.However,most existing skin-mounted electronics use a flexible film with high elastic modulus,which hinders physical activity and causes interfacial delamination and skin irritation.The compliance of hydrogel-based devices can firmly conform to complex,curved surfaces without introducing excessive interfacial stresses.However,most hydrogels still suffer from the weakness of stable and reproducible sensing.In this work,we report a skin-friendly epidermal electronic made of a resilient,self-healing,and recyclable polyvinyl alcohol(PVA)hydrogel.The hydrogel is reinforced through a ternary heterogeneous network for good mechanical robustness while maintaining high stretchability and exceptional conformability.Simultaneously,the abundant dynamic hydrogen bonds give the hydrogel rapid self-healing ability.The assembled hydrogel epidermal electronic is able to stably monitor multiple physiological signals as well as sense the strain level of the skin motion and joint bending.The unique,versatile,environmental and biological friendly epidermal electronics will have broad applications in health care,human-machine interface,augmented reality,and so on.展开更多
Inspired by the fast,agile movements of insects,we present a 1.9 g,4.5 cm in length,piezoelectrically driven,quadrupedal microrobot.This microrobot uses a novel spatial parallel mechanism as its hip joint,which consis...Inspired by the fast,agile movements of insects,we present a 1.9 g,4.5 cm in length,piezoelectrically driven,quadrupedal microrobot.This microrobot uses a novel spatial parallel mechanism as its hip joint,which consists of two spatially orthogonal slider-crank linkages.This mechanism maps two inputs of two independent actuators to the decoupled swing and lift outputs of a leg,and each leg can produce the closed trajectories in the sagittal plane necessary for robot motion.Moreover,the kinematics of the transmission are analyzed,and the parameters of the flexure hinges are designed based on geometrical constraints and yield conditions.The hip joints,legs and exoskeletons are integrated into a five-layer standard laminate for monolithic fabrication which is composed of two layers of carbon fiber,two layers of acrylic adhesive and a polyimide film.The measured output force(15.97 mN)of each leg is enough to carry half of the robot’s weight,which is necessary for the robot to move successfully.It has been proven that the robot can successfully perform forward and turning motions.Compared to the microrobot fabricated with discrete components,the monolithically fabricated microrobot is more capable of maintaining the original direction of locomotion when driven by a forward signal and has a greater speed,whose maximum speed is 25.05 cm/s.展开更多
Parkinson’s disease patients suffer from disorders of speech.The most frequently reported speech problems are weak,hoarse,nasal or monotonous voice,imprecise articulation,slow or fast speech,difficulty starting speec...Parkinson’s disease patients suffer from disorders of speech.The most frequently reported speech problems are weak,hoarse,nasal or monotonous voice,imprecise articulation,slow or fast speech,difficulty starting speech,impaired stress or rhythm,stuttering,and tremor.To improve the speech quality and assist the patient with speech rehabilitation therapy,we have proposed the speech recognition model for Parkinson’s disease patients using transfer learning technique(PSTL),where we have pre-trained the long short-term memory(LSTM)neural network model with our developed publicly available dataset that has been obtained from healthy people through the social media platform.Then,we applied the transfer learning technique to improve the performance of the PSTL framework.The frequency spectrogram masking data augmentation method has been used to alleviate the over-fitting problem so that the word error rate(WER)is further reduced.Even with a limited dataset,our proposed model has effectively reduced the WER from 58% to 44.5% on the original speech dataset and 53.1% to 43% on the denoised speech dataset,which demonstrated the feasibility of our framework.展开更多
Removing different types of artifacts from the electroencephalography(EEG)recordings is a critical step in performing EEG signal analysis and diagnosis.Most of the existing algorithms aim for removing single type of a...Removing different types of artifacts from the electroencephalography(EEG)recordings is a critical step in performing EEG signal analysis and diagnosis.Most of the existing algorithms aim for removing single type of artifacts,leading to a complex system if an EEG recording contains different types of artifacts.With the advancement in wearable technologies,it is necessary to develop an energy-efficient algorithm to deal with different types of artifacts for single-channel wearable EEG devices.In this paper,an automatic EEG artifact removal algorithm is proposed that effectively reduces three types of artifacts,i.e.,ocular artifact(OA),transmission-line/harmonic-wave artifact(TA/HA),and muscle artifact(MA),from a single-channel EEG recording.The effectiveness of the proposed algorithm is verified on both simulated noisy EEG signals and real EEG from CHB-MIT dataset.The experimental results show that the proposed algorithm effectively suppresses OA,MA and TA/HA from a single-channel EEG recording as well as physical movement artifact.展开更多
As a kind of physical signals that could be easily acquired in daily life,photoplethysmography(PPG)signal becomes a promising solution to biometric identification for daily access management system(AMS).State-of-the-a...As a kind of physical signals that could be easily acquired in daily life,photoplethysmography(PPG)signal becomes a promising solution to biometric identification for daily access management system(AMS).State-of-the-art PPG-based identification systems are susceptible to the form of motions and physical conditions of the subjects.In this work,to exploit the advantage of deep learning,we developed an improved deep convolutional neural network(CNN)architecture by using the Gram matrix(GM)technique to convert time-serial PPG signals to two-dimensional images with a temporal dependency to improve accuracy under different forms of motions.To ensure a fair evaluation,we have adopted cross-validation method and“training and testing”dataset splitting method on the TROIKA dataset collected in ambulatory conditions.As a result,the proposed GM-CNN method achieved accuracy improvement from 69.5%to 92.4%,which is the best result in terms of multi-class classification compared with state-of-the-art models.Based on average five-fold cross-validation,we achieved an accuracy of 99.2%,improved the accuracy by 3.3%compared with the best existing method for the binary-class.展开更多
As a primary computation unit,a processing element(PE)is key to the energy efficiency of a convolutional neural network(CNN)accelerator.Taking advantage of the inherent error tolerance of CNNs,approximate computing wi...As a primary computation unit,a processing element(PE)is key to the energy efficiency of a convolutional neural network(CNN)accelerator.Taking advantage of the inherent error tolerance of CNNs,approximate computing with high hardware efficiency has been considered for implementing the computation units of CNN accelerators.However,individual approximate designs such as multipliers and adders can only achieve limited accuracy and hardware improvements.In this paper,an approximate PE is dedicatedly devised for CNN accelerators by synergistically considering the data representation,multiplication and accumulation.An approximate data format is defined for the weights using stochastic rounding.This data format enables a simple implementation of multiplication by using small lookup tables,an adder and a shifter.Two approximate accumulators are further proposed for the product accumulation in the PE.Compared with the exact 8-bit fixed-point design,the proposed PE saves more than 29%and 20%in power-delay product for 3×3 and 5×5 sum of products,respectively.Also,compared with the PEs consisting of state-of-the-art approximate multipliers,the proposed design shows significantly smaller error bias with lower hardware overhead.Moreover,the application of the approximate PEs in CNN accelerators is analyzed by implementing a multi-task CNN for face detection and alignment.We conclude that 1)an approximate PE is more effective for face detection than for alignment,2)an approximate PE with high statistically-measured accuracy does not necessarily result in good quality in face detection,and 3)properly increasing the number of PEs in a CNN accelerator can improve its power and energy efficiency.展开更多
This paper presents a dynamic optical phantom for the simulation of metabolic activities in the brain,and a linear equivalent model is built for control voltage versus substance concentration.A solid-solid dynamic opt...This paper presents a dynamic optical phantom for the simulation of metabolic activities in the brain,and a linear equivalent model is built for control voltage versus substance concentration.A solid-solid dynamic optical phantom is realized by using liquid crystal film as a voltage-controlled light intensity regulator on the surface of basic phantom,which uses epoxy resin as matrix material and nanometer carbon powder and titanium dioxide powder as absorption and scattering dopants,respectively.The dynamic phantom could mimic near-infrared spectrum(NIRS)signals with sampling rate up to 10 Hz,and the maximum simulation errors for oxy-hemoglobin and deoxy-hemoglobin concentrations varying in the range of 1μmol/l are 7.0%and 17.9%,respectively.Compared with similar solid biomimetic phantoms,the adjustable mimic substance concentration range is extended by an order of magnitude,which meets the simulation requirements of most brain NIRS signals.展开更多
基金supported in part by the Open Fund of State Key Laboratory of Integrated Chips and Systems,Hehai Universityin part by the National Science Foundation of China under Grant No.62304133 and No.62350610271.
文摘Reducing the process variation is a significant concern for resistive random access memory(RRAM).Due to its ultrahigh integration density,RRAM arrays are prone to lithographic variation during the lithography process,introducing electrical variation among different RRAM devices.In this work,an optical physical verification methodology for the RRAM array is developed,and the effects of different layout parameters on important electrical characteristics are systematically investigated.The results indicate that the RRAM devices can be categorized into three clusters according to their locations and lithography environments.The read resistance is more sensitive to the locations in the array(~30%)than SET/RESET voltage(<10%).The increase in the RRAM device length and the application of the optical proximity correction technique can help to reduce the variation to less than 10%,whereas it reduces RRAM read resistance by 4×,resulting in a higher power and area consumption.As such,we provide design guidelines to minimize the electrical variation of RRAM arrays due to the lithography process.
基金supported by the National Natural Science Foundation of China(Grant No.60776034)the Open Fund of Key Laboratory of Complex Electromagnetic Environment Science and Technology,China Academy of Engineering Physics(Grant No.2015-0214.XY.K)
文摘The instantaneous reversible soft logic upset induced by the electromagnetic interference（EMI） severely affects the performances and reliabilities of complementary metal–oxide–semiconductor（CMOS） inverters. This kind of soft logic upset is investigated in theory and simulation. Physics-based analysis is performed, and the result shows that the upset is caused by the non-equilibrium carrier accumulation in channels, which can ultimately lead to an abnormal turn-on of specific metal–oxide–semiconductor field-effect transistor（MOSFET） in CMOS inverter. Then a soft logic upset simulation model is introduced. Using this model, analysis of upset characteristic reveals an increasing susceptibility under higher injection powers, which accords well with experimental results, and the influences of EMI frequency and device size are studied respectively using the same model. The research indicates that in a range from L waveband to C waveband, lower interference frequency and smaller device size are more likely to be affected by the soft logic upset.
基金financially supported by the Supporting Foundation of the Ministry of Education (Grant 62501040303)the Pre-research Fund (Grants 9140A26020313JW03371, 9140A260204 14JW03412)the New Century Excellent Talents Support Program from the Ministry of Education of China (Grant NCET-10-0583)
文摘Hummingbirds have a unique way of hover- ing. However, only a few published papers have gone into details of the corresponding three-dimensional vortex struc- tures and transient aerodynamic forces. In order to deepen the understanding in these two realms, this article presents an integrated computational fluid dynamics study on the hovering aerodynamics of a rufous hummingbird. The original morphological and kinematic data came from a former researcher＇s experiments. We found that conical and sta- ble leading-edge vortices （LEVs） with spanwise flow inside their cores existed on the hovering hummingbird＇s wing surfaces. When the LEVs and other near-field vortices were all shed into the wake after stroke reversals, periodically shed bilateral vortex rings were formed. In addition, a strong downwash was present throughout the flapping cycle. Time histories of lift and drag were also obtained. Combining the three-dimensional flow field and time history of lift, we believe that high lift mechanisms （i.e., rotational circulation and wake capture） which take place at stroke reversals in insect flight was not evident here. For mean lift throughout a whole cycle, it is calculated to be 3.60 g （104.0 % of the weight support）. The downstroke and upstroke provide 64.2 % and 35.8 % of the weight support, respectively.
基金supported by the Science and Technology Innovation Council of Shenzhen (KQTD20170810105439418 and JCYJ20200109114237902)the National Natural Science Foundation of China (61903317)+2 种基金the joint funding program of Guangdong Department of Science and Technology and Hongkong Innovation and Technology (2021A0505110015)the Guangdong Basic and Applied Basic Research Foundation (2021B1515420005)the funding from the Shenzhen Institute of Artificial Intelligence and Robotics for Society (AC01202101011 and AC01202101106).
文摘Epidermal sensing devices,which mimic functionalities and mechanical properties of natural skin,offer great potential for real-time health monitoring via continuous checking of vital signs.However,most existing skin-mounted electronics use a flexible film with high elastic modulus,which hinders physical activity and causes interfacial delamination and skin irritation.The compliance of hydrogel-based devices can firmly conform to complex,curved surfaces without introducing excessive interfacial stresses.However,most hydrogels still suffer from the weakness of stable and reproducible sensing.In this work,we report a skin-friendly epidermal electronic made of a resilient,self-healing,and recyclable polyvinyl alcohol(PVA)hydrogel.The hydrogel is reinforced through a ternary heterogeneous network for good mechanical robustness while maintaining high stretchability and exceptional conformability.Simultaneously,the abundant dynamic hydrogen bonds give the hydrogel rapid self-healing ability.The assembled hydrogel epidermal electronic is able to stably monitor multiple physiological signals as well as sense the strain level of the skin motion and joint bending.The unique,versatile,environmental and biological friendly epidermal electronics will have broad applications in health care,human-machine interface,augmented reality,and so on.
基金supported by the Shanghai professional technology service platform under Grant 19DZ2291103.
文摘Inspired by the fast,agile movements of insects,we present a 1.9 g,4.5 cm in length,piezoelectrically driven,quadrupedal microrobot.This microrobot uses a novel spatial parallel mechanism as its hip joint,which consists of two spatially orthogonal slider-crank linkages.This mechanism maps two inputs of two independent actuators to the decoupled swing and lift outputs of a leg,and each leg can produce the closed trajectories in the sagittal plane necessary for robot motion.Moreover,the kinematics of the transmission are analyzed,and the parameters of the flexure hinges are designed based on geometrical constraints and yield conditions.The hip joints,legs and exoskeletons are integrated into a five-layer standard laminate for monolithic fabrication which is composed of two layers of carbon fiber,two layers of acrylic adhesive and a polyimide film.The measured output force(15.97 mN)of each leg is enough to carry half of the robot’s weight,which is necessary for the robot to move successfully.It has been proven that the robot can successfully perform forward and turning motions.Compared to the microrobot fabricated with discrete components,the monolithically fabricated microrobot is more capable of maintaining the original direction of locomotion when driven by a forward signal and has a greater speed,whose maximum speed is 25.05 cm/s.
基金the National Key Research and Development Program of China(No.2019YFB2204500)the Science,Technology and Innovation Action Plan of Shanghai Municipality(No.1914220370)。
文摘Parkinson’s disease patients suffer from disorders of speech.The most frequently reported speech problems are weak,hoarse,nasal or monotonous voice,imprecise articulation,slow or fast speech,difficulty starting speech,impaired stress or rhythm,stuttering,and tremor.To improve the speech quality and assist the patient with speech rehabilitation therapy,we have proposed the speech recognition model for Parkinson’s disease patients using transfer learning technique(PSTL),where we have pre-trained the long short-term memory(LSTM)neural network model with our developed publicly available dataset that has been obtained from healthy people through the social media platform.Then,we applied the transfer learning technique to improve the performance of the PSTL framework.The frequency spectrogram masking data augmentation method has been used to alleviate the over-fitting problem so that the word error rate(WER)is further reduced.Even with a limited dataset,our proposed model has effectively reduced the WER from 58% to 44.5% on the original speech dataset and 53.1% to 43% on the denoised speech dataset,which demonstrated the feasibility of our framework.
基金the National Natural Science Foundation of China(No.61874171)the Alibaba Innovative Research Program of Alibaba Group。
文摘Removing different types of artifacts from the electroencephalography(EEG)recordings is a critical step in performing EEG signal analysis and diagnosis.Most of the existing algorithms aim for removing single type of artifacts,leading to a complex system if an EEG recording contains different types of artifacts.With the advancement in wearable technologies,it is necessary to develop an energy-efficient algorithm to deal with different types of artifacts for single-channel wearable EEG devices.In this paper,an automatic EEG artifact removal algorithm is proposed that effectively reduces three types of artifacts,i.e.,ocular artifact(OA),transmission-line/harmonic-wave artifact(TA/HA),and muscle artifact(MA),from a single-channel EEG recording.The effectiveness of the proposed algorithm is verified on both simulated noisy EEG signals and real EEG from CHB-MIT dataset.The experimental results show that the proposed algorithm effectively suppresses OA,MA and TA/HA from a single-channel EEG recording as well as physical movement artifact.
基金the National Key R&D Program of China(No.2019YFB2204500)the Translational Medicine Cross Research Fund of Shanghai Jiao Tong University(No.ZH2018QNB22)。
文摘As a kind of physical signals that could be easily acquired in daily life,photoplethysmography(PPG)signal becomes a promising solution to biometric identification for daily access management system(AMS).State-of-the-art PPG-based identification systems are susceptible to the form of motions and physical conditions of the subjects.In this work,to exploit the advantage of deep learning,we developed an improved deep convolutional neural network(CNN)architecture by using the Gram matrix(GM)technique to convert time-serial PPG signals to two-dimensional images with a temporal dependency to improve accuracy under different forms of motions.To ensure a fair evaluation,we have adopted cross-validation method and“training and testing”dataset splitting method on the TROIKA dataset collected in ambulatory conditions.As a result,the proposed GM-CNN method achieved accuracy improvement from 69.5%to 92.4%,which is the best result in terms of multi-class classification compared with state-of-the-art models.Based on average five-fold cross-validation,we achieved an accuracy of 99.2%,improved the accuracy by 3.3%compared with the best existing method for the binary-class.
基金supported in part by the National Natural Science Foundation of China under Grant No.62104127the National Key Research and Development Program of China under Grant No.2022YFB4500200.
文摘As a primary computation unit,a processing element(PE)is key to the energy efficiency of a convolutional neural network(CNN)accelerator.Taking advantage of the inherent error tolerance of CNNs,approximate computing with high hardware efficiency has been considered for implementing the computation units of CNN accelerators.However,individual approximate designs such as multipliers and adders can only achieve limited accuracy and hardware improvements.In this paper,an approximate PE is dedicatedly devised for CNN accelerators by synergistically considering the data representation,multiplication and accumulation.An approximate data format is defined for the weights using stochastic rounding.This data format enables a simple implementation of multiplication by using small lookup tables,an adder and a shifter.Two approximate accumulators are further proposed for the product accumulation in the PE.Compared with the exact 8-bit fixed-point design,the proposed PE saves more than 29%and 20%in power-delay product for 3×3 and 5×5 sum of products,respectively.Also,compared with the PEs consisting of state-of-the-art approximate multipliers,the proposed design shows significantly smaller error bias with lower hardware overhead.Moreover,the application of the approximate PEs in CNN accelerators is analyzed by implementing a multi-task CNN for face detection and alignment.We conclude that 1)an approximate PE is more effective for face detection than for alignment,2)an approximate PE with high statistically-measured accuracy does not necessarily result in good quality in face detection,and 3)properly increasing the number of PEs in a CNN accelerator can improve its power and energy efficiency.
基金supported by the National Key Research and Development Program of China under Grant No.STI2030-Major Projects 2022ZD0208500.
文摘This paper presents a dynamic optical phantom for the simulation of metabolic activities in the brain,and a linear equivalent model is built for control voltage versus substance concentration.A solid-solid dynamic optical phantom is realized by using liquid crystal film as a voltage-controlled light intensity regulator on the surface of basic phantom,which uses epoxy resin as matrix material and nanometer carbon powder and titanium dioxide powder as absorption and scattering dopants,respectively.The dynamic phantom could mimic near-infrared spectrum(NIRS)signals with sampling rate up to 10 Hz,and the maximum simulation errors for oxy-hemoglobin and deoxy-hemoglobin concentrations varying in the range of 1μmol/l are 7.0%and 17.9%,respectively.Compared with similar solid biomimetic phantoms,the adjustable mimic substance concentration range is extended by an order of magnitude,which meets the simulation requirements of most brain NIRS signals.
