Highly Efficient Back‑End‑of‑Line Compatible Flexible Si‑Based Optical Memristive Crossbar Array for Edge Neuromorphic Physiological Signal Processing and Bionic Machine Vision

The emergence of the Internet-of-Things is anticipated to create a vast market for what are known as smart edge devices,opening numerous opportunities across countless domains,including personalized healthcare and advanced robotics.Leveraging 3D integration,edge devices can achieve unprecedented miniaturization while simultaneously boosting processing power and minimizing energy consumption.Here,we demonstrate a back-end-of-line compatible optoelectronic synapse with a transfer learning method on health care applications,including electroencephalogram(EEG)-based seizure prediction,electromyography(EMG)-based gesture recognition,and electrocardiogram(ECG)-based arrhythmia detection.With experiments on three biomedical datasets,we observe the classification accuracy improvement for the pretrained model with 2.93%on EEG,4.90%on ECG,and 7.92%on EMG,respectively.The optical programming property of the device enables an ultralow power(2.8×10^(-13) J)fine-tuning process and offers solutions for patient-specific issues in edge computing scenarios.Moreover,the device exhibits impressive light-sensitive characteristics that enable a range of light-triggered synaptic functions,making it promising for neuromorphic vision application.To display the benefits of these intricate synaptic properties,a 5×5 optoelectronic synapse array is developed,effectively simulating human visual perception and memory functions.The proposed flexible optoelectronic synapse holds immense potential for advancing the fields of neuromorphic physiological signal processing and artificial visual systems in wearable applications.

Ferroelectric-controlled graphene plasmonic surfaces for all-optical neuromorphic vision

Artificial visual systems can recognize desired objects and information from complex environments, and are therefore highly desired for pattern recognition, object detection, and imaging applications. However, state-of-the-art artificial visual systems with high recognition performances that typically consist of electronic devices face the challenges of requiring huge storage space and high power consumption owing to redundant data. Here, we report a terahertz(THz) frequency-selective surface using a graphene split-ring resonator driven by ferroelectric polarization for efficient visual system applications. The downward polarization of the ferroelectric material offers an ultrahigh electrostatic field for doping p-type graphene with an anticipated Fermi level. By optimizing the geometric parameters of the devices and modulating the carrier behaviors of graphene, our plasmonic devices exhibit a tunable spectral response in a range of 1.7–6.0 THz with continuous transmission values. The alloptical neural network using graphene plasmonic surfaces designed in this study exhibited excellent performance in visual preprocessing and convolutional filtering and achieved an ultrahigh recognition accuracy of up to 99.3% in training the Modified National Institute of Standards and Technology(MNIST) handwritten digit dataset. These features demonstrate the great potential of graphene plasmonic devices for future smart artificial vision systems.

Retinal Prosthesis:A Potential Benefit for the Blind

Retinal degenerative diseases may induce the degeneration of outer retina and in turn,blindness.Nevertheless,due to the maintenance of inner retina,the coding and processing of visual neurons responses are still able to be executed naturally.Therefore,an effective retinal prosthesis device may be developed by mimicking the function of outer retina:transferring the visual light into artificial stimulus and delivering the stimulus to the retina aiming to evoke the neural responses.As two main developing directions for current retinal prosthesis,epiretinal(ER)and subretinal(SR)prosthesis are both undergoing experimental stage and possessing advantages and limitations.Further investigations in power supply,biocompatibility,etc.are still required.Additionally,suprachoroidal transretinal stimulation(STS)and neurotransmitter-induced stimulation as some other alternatives in retinal prosthesis are also considered as promising research directions,although they are not mature enough to be applied commercially,either.

用于低能耗人工视觉系统的具有互补光调制和低功耗的双极突触有机/无机异质结晶体管

光电突触晶体管将光传感和突触功能集成到单个器件中,在视觉信息采集、识别、记忆和处理的神经形态计算具有显著的优势.然而,现有光电突触的权重更新主要是基于光刺激和电刺激分别调节突触的兴奋和抑制.这种方式严重限制了器件的处理速度和应用场景.在这项工作中,我们提出了双极突触有机/无机异质结晶体管(BSOIHT),可以有效地模拟光刺激下的双向(兴奋/抑制)突触行为.此外,通过优化电极接触位置以及电极材料,晶体管的载流子注入得到了显著改善,使得突触事件功耗降至2.4 fJ.此外,采用BSOIHT构建的神经形态视觉系统,有效地促进了图像预处理,将识别准确率从44.93%大幅提高到87.01%.这为构建低能耗的人工视觉系统提供了新的途径.

Large-area growth of synaptic heterostructure arrays for integrated neuromorphic visual perception chips

Two-dimensional metal chalcogenides have garnered significant attention as promising candidates for novel neuromorphic synaptic devices due to their exceptional structural and optoelectronic properties.However,achieving large-scale integration and practical applications of synaptic chips has proven to be challenging due to significant hurdles in materials preparation and the absence of effective nanofabrication techniques.In a recent breakthrough,we introduced a revolutionary allopatric defect-modulated Fe_(7)S_(8)@MoS_(2)synaptic heterostructure,which demonstrated remarkable optoelectronic synaptic response capabilities.Building upon this achievement,our current study takes a step further by presenting a sulfurization-seeding synergetic growth strategy,enabling the large-scale and arrayed preparation of Fe_(7)S_(8)@MoS_(2)heterostructures.Moreover,a three-dimensional vertical integration technique was developed for the fabrication of arrayed optoelectronic synaptic chips.Notably,we have successfully simulated the visual persistence function of the human eye with the adoption of the arrayed chip.Our synaptic devices exhibit a remarkable ability to replicate the preprocessing functions of the human visual system,resulting in significantly improved noise reduction and image recognition efficiency.This study might mark an important milestone in advancing the field of optoelectronic synaptic devices,which significantly prompts the development of mature integrated visual perception chips.

Image classification on smart agriculture platforms:Systematic literature review

In recent years,smart agriculture has gained strength due to the application of industry 4.0 technologies in agriculture.As a result,efforts are increasing in proposing artificial vision applications to solvemany problems.However,many of these applications are developed separately.Many academic works have proposed solutions integrating image classification techniques through IoT platforms.For this reason,this paper aims to answer the following research questions:(1)What are themain problems to be solvedwith smart farming IoT platforms that incorporate images?(2)What are the main strategies for incorporating image classification methods in smart agriculture IoT platforms?and(3)What are the main image acquisition,preprocessing,transmission,and classification technologies used in smart agriculture IoT platforms?This study adopts a Systematic Literature Review(SLR)approach.We searched Scopus,Web of Science,IEEE Xplore,and Springer Link databases from January 2018 to July 2022.Fromwhich we could identify five domains corresponding to(1)disease and pest detection,(2)crop growth and health monitoring,(3)irrigation and crop protectionmanagement,(4)intrusion detection,and(5)fruits and plant counting.There are three types of strategies to integrate image data into smart agriculture IoT platforms:(1)classification process in the edge,(2)classification process in the cloud,and(3)classification process combined.The main advantage of the first is obtaining data in real-time,and its main disadvantage is the cost of implementation.On the other hand,the main advantage of the second is the ability to process high-resolution images,and its main disadvantage is the need for high-bandwidth connectivity.Finally,themixed strategy can significantly benefit infrastructure investment,butmostworks are experimental.

Prototype chemical synapse chip for spatially patterned neurotransmitter stimulation of the retina ex vivo

Biomimetic stimulation of the retina with neurotransmitters,the natural agents of communication at chemical synapses,could be more effective than electrical stimulation for treating blindness from photoreceptor degenerative diseases.Recent studies have demonstrated the feasibility of neurotransmitter stimulation by injecting glutamate,a primary retinal neurotransmitter,into the retina at isolated single sites.Here,we demonstrate spatially patterned multisite stimulation of the retina with glutamate,offering the first experimental evidence for applicability of this strategy for translating visual patterns into afferent neural signals.To accomplish pattern stimulation,we fabricated a special microfluidic device comprising an array of independently addressable microports connected to tiny on-chip glutamate reservoirs via microchannels.The device prefilled with glutamate was interfaced with explanted rat retinas placed over a multielectrode array(MEA)with the retinal ganglion cells(RGC)contacting the electrodes and photoreceptor surface contacting the microports.By independently and simultaneously activating a subset of the microports with modulated pressure pulses,small boluses of glutamate were convectively injected at multiple sites in alphabet patterns over the photoreceptor surface.We found that the glutamate-driven RGC responses recorded through the MEA system were robust and spatially laid out in patterns strongly resembling the injection patterns.The stimulations were also highly localized with spatial resolutions comparable to or better than electrical retinal prostheses.Our findings suggest that surface stimulation of the retina with neurotransmitters in pixelated patterns of visual images is feasible and an artificial chemical synapse chip based on this approach could potentially circumvent the limitations of electrical retinal prostheses.

Image processing methods to evaluate tomato and zucchini damage in post-harvest stages

Through the supply chain,the quality or quality change of the products can generate important losses.The quality control in some steps is made manually that supposes a high level of subjectivity,controlling the quality and its evolution using automatic systems can suppose a reduction of the losses.Testing some automatic image analysis techniques in the case of tomatoes and zucchini is the main objective of this study.Two steps in the supply chain are considered,the feeding of the raw products into the handling chain(because low quality generates a reduction of the chain productivity)and the cool storage of the processed products(as the value at the market is reduced).It was proposed to analyze the incoming products at the head the processing line using CCD cameras to detect low quality and/or dirty products(corresponding to specific farmers/suppliers,it should be asked to improve to maintain the productivity of the line).The second stage is analyzing the evolution of the products along the cool chain(storage and transport),the use of an App developed to be use under Android was proposed to substitute the“visual”evaluation used in practice.The algorithms used,including stages of pre-treatment,segmentation,analysis and presentation of the results take account of the short time available and the limited capacity of the batteries.High performance techniques were applied to the homography stage to discard some of the images,resulting in better performance.Also threads and renderscript kernels were created to parallelize the methods used on the resulting images being able to inspect faster the products.The proposed method achieves success rates comparable to,and improving,the expert inspection.