Abstract Neuromorphic engineering aims to advance computing by mimicking the brain's efficient processing, where data is encoded as asynchronous temporal events. This eliminates need for a synchronisation clock and minimises power consumption when no present. However, many benchmarks neuromorphic algorithms primarily focus on spatial features, neglecting dynamics that are inherent most sequence-based tasks. gap may lead evaluations fail fully capture unique strengths characteristics of...

We present an end-to-end trainable modular event-driven neural architecture that uses local synaptic and threshold adaptation rules to perform transformations between arbitrary spatio-temporal spike patterns. The represents a highly abstracted model of existing Spiking Neural Network (SNN) architectures. proposed Optimized Deep Event-driven network Architecture (ODESA) can simultaneously learn hierarchical features at multiple time scales. ODESA performs online learning without the use error...

In recent years, a new generation of low-power, neuromorphic, event-based vision sensors has been gaining popularity for their very low latency and data sparsity. Though the conventional frame-based cameras have advanced in lot ways, they suffer from redundancy temporal latency. The bio-inspired artificial retinas eliminate by capturing only change illumination at each pixel asynchronously communicating binary spikes. this work, we propose system to achieve task human activity recognition...

This paper presents a reconfigurable digital implementation of an event-based binaural cochlear system on Field Programmable Gate Array (FPGA). It consists pair the Cascade Asymmetric Resonators with Fast Acting Compression (CAR-FAC) cochlea models and leaky integrate-and-fire (LIF) neurons. Additionally, we propose event-driven SpectroTemporal Receptive (STRF) Feature Extraction using Adaptive Selection Thresholds (FEAST). is tested TIDIGTIS benchmark compared current auditory signal...

This paper presents an efficient hardware implementation of the recently proposed Optimized Deep Event-driven Spiking Neural Network Architecture (ODESA). ODESA is first network to have end-to-end multi-layer online local supervised training without using gradients and has combined adaptation weights thresholds in hierarchical structure. research shows that architecture can be implemented efficiently on a large scale hardware. The consists (SNN) individual modules for each layer enable...

This paper presents an expansion and evaluation of the hardware architecture for Optimized Deep Event-driven Spiking Neural Network Architecture (ODESA). ODESA is a state-of-the-art, event-driven multi-layer (SNN) that offers end-to-end, online, local supervised training method. In previous work, was successfully implemented on Field-Programmable Gate Array (FPGA) hardware, showcasing its effectiveness in resource-constrained environments. Building upon implementation, this research focuses...

Manifold amount of video data gets generated every minute as we read this document, ranging from surveillance to broadcasting purposes. There are two roadblocks that restrain us using such, first being the storage which restricts only storing information based on hardware constraints. Secondly, computation required process is highly expensive makes it infeasible work them. Compressive sensing(CS)[2] a signal technique[11], through optimization, sparsity can be exploited recover far fewer...

We will demonstrate a real-time implementation of protoobject based neuromorphic visual saliency model [1] on an embedded processing board. Visual models are difficult to implement in hardware for applications due their computational complexity. The conventional is not optimal because the requirement large number convolution operations filtering several feature channels across multiple image pyramids. Our current considers dynamic temporal motion change by convoluting along time efficiently...

Typically a 1-2MP CCTV camera generates around 7-12GB of data per day. Frame-by-frame processing such an enormous amount requires hefty computational resources. In recent years, compressive sensing approaches have shown impressive compression results by reducing the sampling bandwidth. Different mechanisms were developed to incorporate in image and video acquisition. Though all-CMOS [1], [2] sensor cameras that perform can help save lot bandwidth on minimize memory required store videos,...

While there is much focus on hardware advances that accellerate the simulation of large scale spiking neural networks, it worthwhile to shift our attention language may also support accelerated network simulation. Some gains in biologically faithful neuronal can be achieved by applying recent computer features. For example, Julia supports Sparse Compressed Arrays, Static furthermore provides very extensive for CUDA GPU, as well a plethora reduced precision types. high-level syntax...

Scalable methods for representing the transient behaviour of large populations neurons are needed. In this work, we present an algorithm that can detect fine-grained repetitions quickly across spiking datasets. The proposed method enables us to quantify both state and transitions in spike trains. We were motivated create tool because existing tools also reoccurring patterns trains often not scalable or they limited towards detecting only particular types patterns. demonstrate a geometric...

Reinforcement Learning (RL) provides a powerful framework for decision-making in complex environments. However, implementing RL hardware-efficient and bio-inspired ways remains challenge. This paper presents novel Spiking Neural Network (SNN) architecture solving problems with real-valued observations. The proposed model incorporates multi-layered event-based clustering, the addition of Temporal Difference (TD)-error modulation eligibility traces, building upon prior work. An ablation study...

This work explores using a local spike-timing-dependent adaptation of thresholds and weights to learn classify spectro-temporal representations audio represented by neural spikes. We use the Cascade Asymmetric Resonators with Fast Acting Compression (CAR-FAC) cochlea model Leaky Integrated-and-Fire (LIF) neurons generate spikes from audio. event stream is fed into Optimised Deep Event-driven Spiking Architecture (ODESA) features in hierarchical architecture. The cochlear events provide...

Every day around the world, interminable terabytes of data are being captured for surveillance purposes. A typical 1-2MP CCTV camera generates 7-12GB per day. Frame-by-frame processing such enormous amount requires hefty computational resources. In recent years, compressive sensing approaches have shown impressive results in signal by reducing sampling bandwidth. Different mechanisms were developed to incorporate image and video acquisition. Pixel-wise coded exposure is one among promising...

This paper presents a reconfigurable digital implementation of an event-based binaural cochlear system on Field Programmable Gate Array (FPGA). It consists pair the Cascade Asymmetric Resonators with Fast Acting Compression (CAR FAC) cochlea models and leaky integrate fire (LIF) neurons. Additionally, we propose event-driven SpectroTemporal Receptive (STRF) Feature Extraction using Adaptive Selection Thresholds (FEAST). is tested TIDIGTIS benchmark compared current auditory signal processing...