Neuromorphic Signal Research Articles

Hybrid All-in-Focus Imaging From Neuromorphic Focal Stack.

Creating an image focal stack requires multiple shots, which captures images at different depths within the same scene. Such methods are not suitable for scenes undergoing continuous changes. Achieving an all-in-focus image from a single shot poses significant challenges, due to the highly ill-posed nature of rectifying defocus and deblurring from a single image. In this paper, to restore an all-in-focus image, we introduce the neuromorphic focal stack, which is defined as neuromorphic signal streams captured by an event/ a spike camera during a continuous focal sweep, aiming to restore an all-in-focus image. Given an RGB image focused at any distance, we harness the high temporal resolution of neuromorphic signal streams. From neuromorphic signal streams, we automatically select refocusing timestamps and reconstruct corresponding refocused images to form a focal stack. Guided by the neuromorphic signal around the selected timestamps, we can merge the focal stack using proper weights and restore a sharp all-in-focus image. We test our method on two distinct neuromorphic cameras. Experimental results from both synthetic and real datasets demonstrate a marked improvement over existing State-of-the-Art methods.

Hardware-Feasible and Efficient N-Type Organic Neuromorphic Signal Recognition via Reservoir Computing.

Organic electrochemical synaptic transistors (OESTs), inspired by the biological nervous system, have garnered increasing attention due to their multifunctional applications in neuromorphic computing. However, the practical implementation of OESTs for signal recognition-particularly those utilizing n-type organic mixed ionic-electronic conductors (OMIECs)-still faces significant challenges at the hardware level. Here, a state-of-the-art small-molecule n-type OEST integrated within a physically simple and hardware feasible reservoir-computing (RC) framework for practical temporal signal recognition is presented. This integration is achieved by leveraging the adjustable synaptic properties of the n-OEST, which exhibits tunable nonlinear short-term memory, transitioning from volatility to nonvolatility, and demonstrating adaptive temporal specificity. Additionally, the nonvolatile OEST offers 256 conductance levels and a wide dynamic range (≈147) in long-term potentiation/depression (LTP/LTD), surpassing previously reported n-OESTs. By combining volatile n-OESTs as reservoirs with a single-layer perceptron readout composed of nonvolatile n-OEST networks, this physical RC system achieves substantial recognition accuracy for both handwritten-digit images (94.9%) and spoken digit (90.7%), along with ultrahigh weight efficiency. Furthermore, this system demonstrates outstanding accuracy (98.0%) by grouped RC in practical sleep monitoring, specifically in snoring recognition. Here, a reliable pathway for OMIEC-driven computing is presented to advance bioinspired hardware-based neuromorphic computing in the physical world.

Bidirectional Photovoltage-Driven Oxide Transistors for Neuromorphic Visual Sensors.

Biological vision is one of the most important parts of the human perception system. However, emulating biological visuals is challenging because it requires complementary photoexcitation and photoinhibition. Here, the study presents a bidirectional photovoltage-driven neuromorphic visual sensor (BPNVS) that is constructed by monolithically integrating two perovskite solar cells (PSCs) with dual-gate ion-gel-gated oxide transistors. PSCs act as photoreceptors, converting external visual stimuli into electrical signals, whereas oxide transistors generate neuromorphic signal outputs that can be adjusted to produce positive and negative photoresponses. This device mimics the human vision system's ability to recognize colored and color-mixed patterns. The device achieves a static color recognition accuracy of 96% by utilizing the reservoir computing system for feature extraction. The BPNVS mem-reservoir chip is also proposed for handing object movement and dynamic color recognition. This work is a significant step forward in neuromorphic sensing and complex pattern recognition.

Neuromorphic Signal Classification Using Organic Electrochemical Transistor Array and Spiking Neural Simulations

Neuromorphic Signal Classification Using Organic Electrochemical Transistor Array and Spiking Neural Simulations

Bio‐inspired Two‐dimensional Nanofluidic Ionic Transistor for Neuromorphic Signal Processing

AbstractVoltage‐gated ion channels prevalent in neurons play important roles in generating action potential and information transmission by responding to transmembrane potential. Fabricating bio‐inspired ionic transistors with ions as charge carriers will be crucial for realizing neuro‐inspired devices and brain‐liking computing. Here, we reported a two‐dimensional nanofluidic ionic transistor based on a MXene membrane with sub‐1 nm interlayer channels. By applying a gating voltage on the MXene nanofluidic, a transmembrane potential will be generated to active the ionic transistor, which is similar to the transmembrane potential of neuron cells and can be effectively regulated by changing membrane parameters, e.g., thickness, composition, and interlayer spacing. For the symmetric MXene nanofluidic, a high on/off ratio of ~2000 can be achieved by forming an ionic depletion or accumulation zone, contingent on the sign of the gating potential. An asymmetric PET/MXene‐composited nanofluidic transitioned the ionic transistor from ambipolar to unipolar, resulting in a more sensitive gate voltage characteristic with a low subthreshold swing of 560 mV/decade. Furthermore, ionic logic gate circuits, including the “NOT”, “NAND”, and “NOR” gate, were implemented for neuromorphic signal processing successfully, which provides a promising pathway towards highly parallel, low energy consumption, and ion‐based brain‐like computing.

Bio-inspired Two-dimensional Nanofluidic Ionic Transistor for Neuromorphic Signal Processing.

Voltage-gated ion channels prevalent in neurons play important roles in generating action potential and information transmission by responding to transmembrane potential. Fabricating bio-inspired ionic transistors with ions as charge carriers will be crucial for realizing neuro-inspired devices and brain-liking computing. Here, we reported a two-dimensional nanofluidic ionic transistor based on a MXene membrane with sub-1 nm interlayer channels. By applying a gating voltage on the MXene nanofluidic, a transmembrane potential will be generated to active the ionic transistor, which is similar to the transmembrane potential of neuron cells and can be effectively regulated by changing membrane parameters, e.g., thickness, composition, and interlayer spacing. For the symmetric MXene nanofluidic, a high on/off ratio of ~2000 can be achieved by forming an ionic depletion or accumulation zone, contingent on the sign of the gating potential. An asymmetric PET/MXene-composited nanofluidic transitioned the ionic transistor from ambipolar to unipolar, resulting in a more sensitive gate voltage characteristic with a low subthreshold swing of 560 mV/decade. Furthermore, ionic logic gate circuits, including the "NOT", "NAND", and "NOR" gate, were implemented for neuromorphic signal processing successfully, which provides a promising pathway towards highly parallel, low energy consumption, and ion-based brain-like computing.

Single Wavelength Operating Neuromorphic Device Based on a Graphene-Ferroelectric Transistor.

As global data generation continues to rise, there is an increasing demand for revolutionary in-memory computing methodologies and efficient machine learning solutions. Despite recent progress in electrical and electro-optical simulations of machine learning devices, the all-optical nonthermal function remains challenging, with single wavelength operation still elusive. Here we report on an optical and monochromatic way of neuromorphic signal processing for brain-inspired functions, eliminating the need for electrical pulses. Multilevel synaptic potentiation-depression cycles are successfully achieved optically by leveraging photovoltaic charge generation and polarization within the photoferroelectric substrate interfaced with the graphene sensor. Furthermore, the demonstrated low-power prototype device is able to reproduce exact signal profile of brain tissues yet with more than 2 orders of magnitude faster response. The reported properties should trigger all-optical and low power artificial neuromorphic development based on photoferroelectric structures.

Exploring Filter Banks and Spike Interval Statistics of Level-Crossing ADCs for Fault Diagnosis of Rolling Element Bearings

Nowadays, lots of data are generated in industries using vibration sensors to evaluate the equipment’s working condition and identify faults. A significant challenge is that only a small fraction of data can be transmitted for intelligent fault diagnosis and storage. The edge processing capacity is often insufficient for advanced analysis due to time and resource constraints. The neuromorphic signal encoding scheme efficiently reduces the data rate by encoding relevant signal changes into spike trains while discarding redundant information and noise, enabling energy-efficient neuromorphic processing. Due to the presence of dominant operational features and noise in the original measurements, signal pre-processing is required to extract the relevant features before spike coding and processing. The work investigates the effects of different filter banks (pre-processing methods) on the spike encodings for vibration measurements from bearings. This also includes bearing fault features diagnosis based on statistical analysis of generated spikes. The comparative analysis is made for benchmarking different signal pre-processing methods (e.g., envelope, empirical mode decomposition (EMD), and gammatone filter) on bearing vibration datasets. An event-triggered scheme, i.e., Level-crossing analog-to-digital converters (LC-ADCs) is applied to encode the vibration measurement to spikes. Inter-spike intervals (ISIs) statistics are analysed for fault diagnosis of bearings. The results obtained for CWRU bearing databases indicate a possible fault detection and diagnosis with significant data rate reduction and an opportunity for improved computational efficiency. With the developed approach, the envelope filter is found to be the most efficient of all. This work enables a new approach to improve the energy efficiency of condition monitoring systems and further sets a new course of research development in this area using neuromorphic technologies.

78‐1: Invited Paper: Tactile Scanner as Neuromorphic Skin for Computer Haptics

In computer haptics, haptic devices provide sensory feedback that simulates physical properties and forces for recreating a feeling of touch when interacting with a virtual object. To generate such a feeling, a tactile sensor is required to collect tactile information similar to what human hand or skin does and then render it to drive a high‐fidelity artificial skin for tactile display. In this work, a flexible tactile scanner with a TFT‐based neuromorphic circuit as sensor interface is proposed to acquire tactile information with bionic characteristics. In order to recognize and classify different materials and textured surfaces, such a neuromorphic signal conversion is of necessity to generate neuromorphic data that can be used for artificial neuron network training and learning. Further, the following labelled bionic tactile information can be utilized to generate a motor‐neuron‐like signal to drive a high‐fidelity bionic neuromorphic skin for tactile display, enabling to generate a touch feeling of a virtual object in Metaverse. Therefore, as the first step, a tactile scanner as neuromorphic skin must be used to acquire tactile data for such a purpose.

Neuromorphic Signal Filter for Robot Sensoring.

Noise management associated with input signals in sensor devices arises as one of the main problems limiting robot control performance. This article introduces a novel neuromorphic filter model based on a leaky integrate and fire (LIF) neural model cell, which encodes the primary information from a noisy input signal and delivers an output signal with a significant noise reduction in practically real-time with energy-efficient consumption. A new approach for neural decoding based on the neuron-cell spiking frequency is introduced to recover the primary signal information. The simulations conducted on the neuromorphic filter demonstrate an outstanding performance of white noise rejecting while preserving the original noiseless signal with a low information loss. The proposed filter model is compatible with the CMOS technology design methodologies for implementing low consumption smart sensors with applications in various fields such as robotics and the automotive industry demanded by Industry 4.0.

Efficient Neuromorphic Signal Processing with Resonator Neurons

Efficient Neuromorphic Signal Processing with Resonator Neurons

Non-spectroscopic sensing enabled by an electro-optical reservoir computer

A fully bio-inspired apparatus to perform non-spectroscopic sensing to discriminate and determine the constituent concentrations of a chemical mixture is proposed. Here, fully bio-inspired means that it is comprised of a biomimetic sensor and a neuromorphic signal processor. The sensor is similar to the human eye in terms of its trichromat architecture and overlapping spectral response. The information processor is a neuromorphic system based on an electro-optical implementation of a reservoir computer. A time-stepping signal algorithm based on the Z -bilinear transformation has been developed to realistically simulate the electro-optical reservoir computer; this demonstrates the discrimination and chemical concentration determination tasks. We believe such an apparatus offers potential benefits in areas in which chemical composition needs to be monitored in real time, for example in chemical processing, and food-beverage (fermentation/brewing) and environmental monitoring.

A Native SPICE Implementation of Memristor Models for Simulation of Neuromorphic Analog Signal Processing Circuits

Since the memristor emerged as a programmable analog storage device, it has stimulated research on the design of analog/mixed-signal circuits with the memristor as the enabler of in-memory computation. Due to the difficulty in evaluating the circuit-level nonidealities of both memristors and CMOS devices, SPICE-accuracy simulation tools are necessary for perfecting the art of neuromorphic analog/mixed-signal circuit design. This article is dedicated to a native SPICE implementation of the memristor device models published in the open literature and develops case studies of applying such a circuit simulation with MOSFET models to study how device-level imperfections can make adversarial effects on the analog circuits that implement neuromorphic analog signal processing. Methods on memristor stamping in the framework of modified nodal analysis formulation are presented, and implementation results are reported. Furthermore, functional simulations on neuromorphic signal processing circuits including memristors and CMOS devices are carried out to validate the effectiveness of the native SPICE implementation of memristor models from the perspectives of simulation accuracy, efficiency, and convergence for large-scale simulation tasks.

Multi-channel optical neuromorphic processor for frequency-multiplexed signals

Here we develop the first optical neuromorphic processor for frequency-multiplexed and multi-channel signals and demonstrate its application for orthogonal frequency-division multiplexing and wavelength-division multiplexing. The presented architecture supports multichannel operation through incorporation of the optical Fourier transform and dispersion-managed fiber echo state network analogue enabling processing of dual-quadrature high bandwidth signals. We demonstrate applicability of the design for prediction and equalization tasks. The proposed technology paves the way to multi-channel neuromorphic signal processing.

Artificial tactile peripheral nervous system supported by self-powered transducers

The tactile peripheral nervous system innervating human hands, which is essential for sensitive haptic exploration and dexterous object manipulation, features overlapped receptive fields in the skin, arborization of peripheral neurons and many-to-many synaptic connections. Inspired by the structural features of the natural system, we report a supersensitive artificial slowly adapting tactile afferent nervous system based on the triboelectric nanogenerator technology. Using tribotronic transistors in the design of mechanoreceptors, the artificial afferent nervous system exhibits the typical adapting behaviours of the biological counterpart in response to mechanical stimulations. The artificial afferent nervous system is self-powered in the transduction and event-driven in the operation. Moreover, it has inherent proficiency of neuromorphic signal processing, delivering a minimum resolvable dimension two times smaller than the inter-receptor distance which is the lower limit of the dimension that existing electronic skins can resolve. These results open up a route to scalable neuromorphic skins aiming at the level of human’s exceptional perception for neurorobotic and neuroprosthetic applications.

Dispersion-Managed Fiber Echo State Network Analogue With High (Including THz) Bandwidth

We propose a design for high (including THz) bandwidth neuromorphic signal processing based on fiber echo state network analogue and demonstrate through numerical modeling its efficiency for distortion mitigation in optical communications with high-order (including 1024-QAM) formats. This is achieved via all-optical implementation of masking and neural functionality by utilizing dispersion and nonlinear properties of the fiber. The design is flexible and format-transparent making it relevant for future communication systems. The model is simple to implement, as it requires only two attenuators and pumps in addition to a traditional nonlinear optical loop mirror (NOLM), and offers a vast range of optimization parameters.

Neuromorphic Functions of Light in Parity-Time-Symmetric Systems.

As an elementary processor of neural networks, a neuron performs exotic dynamic functions, such as bifurcation, repetitive firing, and oscillation quenching. To achieve ultrafast neuromorphic signal processing, the realization of photonic equivalents to neuronal dynamic functions has attracted considerable attention. However, despite the nonconservative nature of neurons due to energy exchange between intra‐ and extra‐cellular regions through ion channels, the critical role of non‐Hermitian physics in the photonic analogy of a neuron has been neglected. Here, a neuromorphic non‐Hermitian photonic system ruled by parity‐time symmetry is presented. For a photonic platform that induces the competition between saturable gain and loss channels, dynamical phases are classified with respect to parity‐time symmetry and stability. In each phase, unique oscillation quenching functions and nonreciprocal oscillations of light fields are revealed as photonic equivalents of neuronal dynamic functions. The proposed photonic system for neuronal functionalities will become a fundamental building block for light‐based neural signal processing.

Ultra-fast artificial neuron: generation of picosecond-duration spikes in a current-driven antiferromagnetic auto-oscillator

We demonstrate analytically and numerically, that a thin film of an antiferromagnetic (AFM) material, having biaxial magnetic anisotropy and being driven by an external spin-transfer torque signal, can be used for the generation of ultra-short “Dirac-delta-like” spikes. The duration of the generated spikes is several picoseconds for typical AFM materials and is determined by the inplane magnetic anisotropy and the effective damping of the AFM material. The generated output signal can consist of a single spike or a discrete group of spikes (“bursting”), which depends on the repetition (clock) rate, amplitude, and shape of the external control signal. The spike generation occurs only when the amplitude of the control signal exceeds a certain threshold, similar to the action of a biological neuron in response to an external stimulus. The “threshold” behavior of the proposed AFM spike generator makes possible its application not only in the traditional microwave signal processing but also in the future neuromorphic signal processing circuits working at clock frequencies of tens of gigahertz.

Asynchronous spiking photonic neuron for lightwave neuromorphic signal processing

We developed an asynchronous spiking photonic neuron that forms the basic building block for hybrid analog/digital lightwave neuromorphic processing. Our approach enables completely asynchronous spiking in response to input signals while maximizing the throughput relative to synchronous approaches. Asynchronous operation is achieved by generating the spike source for the photonic neuron through four-wave mixing. This hybrid analog/digital photonic neuron has an electro-absorption modulator as the temporal integration unit for analog processing, while the digital processing portion employs optical thresholding in a highly Ge-doped nonlinear loop mirror.

Signal feature recognition based on lightwave neuromorphic signal processing

We developed a hybrid analog/digital lightwave neuromorphic processing device that effectively performs signal feature recognition. The approach, which mimics the neurons in a crayfish responsible for the escape response mechanism, provides a fast and accurate reaction to its inputs. The analog processing portion of the device uses the integration characteristic of an electro-absorption modulator, while the digital processing portion employ optical thresholding in a highly Ge-doped nonlinear loop mirror. The device can be configured to respond to different sets of input patterns by simply varying the weights and delays of the inputs. We experimentally demonstrated the use of the proposed lightwave neuromorphic signal processing device for recognizing specific input patterns.