Search

Weightless neural networks (WNNs) are an alternative pattern recognition technique where RAM nodes function as neurons. As both training and inference require mostly table lookups, few additions, and no multiplications, WNNs are suitable for high-performance and low-power embedded applications. This work introduces a novel approach to implement WiSARD, the leading WNN state-of-the-art architecture, completely eliminating memories and arithmetic circuits and utilizing only logic functions. The approach creates compressed minimized implementations by converting trained WNN nodes from lookup tables to logic functions. The proposed LogicWiSARD is implemented in FPGA and ASIC technologies to illustrate its suitability for edge inference. Experimental results show more than 80% reduction in energy consumption when the proposed LogicWiSARD model is compared with a multilayer perceptron network (MLP) of equivalent accuracy. Compared to previous work on FPGA implementations for WNNs, convolutional neural networks, and binary neural networks, the energy savings of LogicWiSARD range between 32.2% and 99.6%. info:eu-repo/semantics/acceptedVersion

With the prevalence of deep learning (DL) in many applications, researchers are investigating different ways of optimizing FPGA architecture and CAD to achieve better quality-of-results (QoR) on DL-based workloads. In this optimization process, benchmark circuits are an essential component; the QoR achieved on a set of benchmarks is the main driver for architecture and CAD design choices. However, current academic benchmark suites are inadequate, as they do not capture any designs from the DL domain. This work presents a new suite of DL acceleration benchmark circuits for FPGA architecture and CAD research, called Koios. This suite of 19 circuits covers a wide variety of accelerated neural networks, design sizes, implementation styles, abstraction levels, and numerical precisions. These designs are larger, more data parallel, more heterogeneous, more deeply pipelined, and utilize more FPGA architectural features compared to existing open-source benchmarks. This enables researchers to pin-point architectural inefficiencies for this class of workloads and optimize CAD tools on more realistic benchmarks that stress the CAD algorithms in different ways. In this paper, we describe the designs in our benchmark suite, present results of running them through the Verilog-to-Routing (VTR) flow using a recent FPGA architecture model, and identify key insights from the resulting metrics. On average, our benchmarks have 3.7x more netlist primitives, 1.8x and 4.7x higher DSP and BRAM densities, and 1.7x higher frequency with 1.9x more near-critical paths compared to the widely-used VTR suite. Finally, we present two example case studies showing how architectural exploration for DL-optimized FPGAs can be performed using our new benchmark suite.

The article presents a technique to estimate the force transmitted in pneumatic artificial muscle based assistive devices by sensorizing the actuator. For this purpose, the development of a conductive liquid metal based elastomeric strain sensor has been discussed with the required characterization of its performance. A comparative analysis is shown between the modes of attachment of the developed sensor on the actuator so as to obtain a proper range of response from the sensor. A two-step algorithm has been implemented to estimate the force transmitted in series from an actuator used to assist the human knee and ankle during walking. The estimation results are validated against the readings from an actual inline loadcell, showing good agreement with one another.

Researchers have developed various different static models for Pneumatic artificial muscle actuators using virtual work technique by using the volume of air entering into the actuator. A comparative study of the measurement and estimation of the air consumption of a pneumatic artificial muscle actuator during its actuation is performed in this article. Such actuators being non-symmetrical, require an extensive hardware to accurately measure their volumes by water displacement technique. The relative performance of two alternative methods, blob analysis and flow rate measurement are analyzed and the anomalies have been reasoned out. A short study on the identification of the actuator inflation process has been performed to show how the estimation technique could be improved.

Background and Aim: Handwritten Digit Recognition has a wide variety of applications in postal mail order, phone records search, automatic car number plate recognition, and in the medical sector that observed how Machine Learning makes the daily tasks simpler and more efficient. This paper aims to improve the classification accuracy of existing handwritten digit systems, thus improve their efficiency. Methodology: The proposed system consists of an enhanced decision function by adding a “Bias Probability” Function. The function adds negative weights to the output classes (0-9) that have a high positive bias and add a positive weight to the output classes that have a high negative bias to neutralize the effect of this high negative bias. Therefore, the Bayesian Classifier function has been enhanced thereby improving the accuracy of classification, which will further improve the performance of the multiclass probability categorization. Result: An increase of 5.6% was observed in the overall accuracy of handwritten digit classification using the Modified National Institute of Standards and Technology (MNIST) Dataset. Conclusion: From the results, it is clear that the proposed system, enhances the main decision function to further improve the accuracy with no significant increase in the processing time.

Stretchable conductive textiles are widely used in the customizable flexible strain, force and pressure sensors for developing highly compliant components in bio-robotics. The article presents the characterization process of such a stretchable conductive fabric. The Pneumatic artificial muscles, viable lightweight and compliant alternatives to conventional actuators that require specialized sensors to monitor their axial contraction according to the degree of miniaturization. In order to avoid an arduous process of fabrication involved in conductive liquid metal-based sensors, an alternative economical and easily developable sensing system is introduced. Based upon the results of the characterization process, the appropriate sensing elements are chosen to develop a serial sensing element-based sensor with a good gauge factor. Such sensors have potential applications in other areas like human joint monitoring, soft robotics, etc. where a good gauge factor is required to maintain the same form factor, without inhibiting any natural motion of the user.

Designing efficient hardware for accelerating artificial intelligence (AI) and machine learning (ML) applications is a major challenge. Rapidly changing algorithms and neural network architectures make FPGA based designs an attractive solution. But the generic building blocks available in current FPGAs (Logic Blocks (LBs), multipliers, DSP blocks) limit the acceleration that can be achieved. We propose Hamamu, a modification to the current FPGA architecture that makes FPGAs specialized for ML applications. Specifically, we propose adding hard matrix multiplier blocks (matmuls) into the FPGA fabric. These matmuls are implemented using systolic arrays of MACs (Multiply-And-Accumulate) and can be connected using programmable direct interconnect between neighboring matmuls to make larger systolic matrix multipliers. We explore various matmul sizes ($2\times 2\times 2$, $4\times 4\times 4$, $8\times 8\times 8$, $16\times 16\times 16$) and various strategies to place these blocks on the FPGA (Columnar, Surround, Hybrid). We find that providing $4\times 4\times 4$ hard matrix multiplier blocks in an FPGA speeds up neural networks from MLPerf benchmarks by up to $\sim 3.9x$, compared to a Stratix-10 like FPGA with equal number of MACs, same MAC architecture and high DSP:LB ratio. Although the flexibility of the FPGA will reduce for non-ML applications, an FPGA with hard matrix multipliers is a faster, and more area efficient hardware accelerator for ML applications, compared to current FPGAs.

A pneumatic haptic display for collaborative robotics applications is presented within this paper. The 3×3 tactor array consists of a polymeric parallelepiped with 9 cylindrical chambers (3mm diameter) providing haptic feedback by means of pneumatic actuation. A preliminary study, on four subjects, has been performed to identify the optimal pressure to deliver the tactile stimulation and assess system functionality. Four different spatial stimuli, based on the activation of different chambers of the tactor array, were provided at three different pressure levels (35, 70, 105 kPa). Results suggest that a pressure of 70 kPa leads to a detection accuracy of 96.9%. Further studies will focus on a deeper assessment of the tactile display using different configurations.

Photoplethysmography(ppg) is an unadorned optical technique. It is a technique for acquiring plethysmogram optically that helps in diagnosing the changes in the volume of blood within an organ in the whole body by detecting blood volume change in micro vascular bed of tissue that further helps to monitor/ analyze heart rate, depth of anesthesia, cardiac cycle and other circulatory functions. It works on the theory of transmission and reflection of green light through the skin as every pulse pressure affects the amount of green light that is transmitted and reflected. PPG is categorized into two types, one is done by having contact with the skin that is generally done using fitness bands, smartwatches and pulse oximeter and another one is contact less that is done via high-quality cameras that detects the changes in absorption and reflection of green light from skin that is not visible with the naked eyes. This technique of generating plethysmogram without any contact with the skin or through cameras is also known as Remote photoplethysmography (rPPG). Each peak in photoplethysmograph will represent a cardiac cycle. Keywords - Photoplethysmography, Artificial Intelligence, plethysmogram, Security, Biomedical Monitoring, Correlation, Reliability

Every Technology in the market has some flaws and some advantages. Similar is the case of IoT (that is known as Internet of things). As its name suggests Internet of Things means transferring data over network in between connected devices without any human interaction and providing a promising environment for the actions that are to be performed. In IoT environment human interaction is bare minimum so chances of human error are also bare minimum which will help in increasing accuracy in various applications where humans are not able to perform well, machines can perform much better. IoT nowadays has become a basic necessity for humans and in future it will be going to become as essential for everyone as water is for humans to survive. In this paper topic of discussion is about various applications of IoT in various sectors, It's Security and privacy concerns which will throw the light on how various attacks and flaws could be prevented and then future research trends will be discussed.

With the increasing concerns about the security of data, Blockchain a decentralised public ledger, linked block structure which verifies and stores data and follows trusted consensus algorithms which ensure synchronisation of data in a distributed peer to peer network acquired a lot of attention as well as discussions recently. As it makes sure a tamper-proof digital platform for decision-making processes to store and share data among different individuals, authorities and organisations who do not trust each other. This technology is considered as the future of data security and integrity and is expected to serve in a lot of new or existing potential application, so the motive of this paper is to present insights of blockchain's architecture, it's security background, and it elaborate the two versions of blockchain: permissioned and permission-less. and some of its applications. Moreover, the first decentralized cryptocurrency and a public ledger named bitcoin are also elaborated in this paper. Furthermore, we summarize some typical implementations in blockchain, the lifecycle of bitcoin and explore future research challenges that still need to be addressed in order to preserve privacy when blockchain is used.

This study aimed at the design and fabrication of a modular soft robotic manipulator for use in Minimally Invasive Surgeries(MIS). A single segment of such manipulator was fabricated with soft silicone polymer and three Shape Memory Alloy (SMA) spring wires connected 120 apart were used as actuators. Prilimary experiments conducted on a couple of such segments connected one over the other showed very promising results. A single segment with only one wire actuated showed a bending angle of 41 while two segments connected end to end showed a bending angle of 84. The prototype is an improvement over all existing manipulation systems used in MIS in terms of bending angle, actuation time and recovery time and is capable of attaining complex orientations in 3D space.

Advancement of technology and rapid Internet made information to travel over the world in an easy and economical way. This has made people to anguish about their privacy. Steganography is a technique that thwarts unauthorized users to have access to the crucial data. There are various steganographic methods that users can use to hide and mix information within other information that make it arduous to recognize by unauthorised users. This paper gives an overview of LSB technique of steganography and further we proposed and implemented a unique algorithm to base steganography which is an enhanced and improved technique better in all aspects. The paper further compares our proposed enhanced technique with the existing LSB technique on various parameters to show effective and satisfactory results.

Production chains featuring industrial vision are becoming more and more widespread. Those processes are often heavy and applied to high definition images with important frame rate. Powerful calculators are thus needed to follow the ever growing production rate. NVIDIA is currently designing interfaces providing a CUDA (Compute Unified Device Architecture) architecture allowing parallel data computation. This could increase the performance of every operating system using graphical processing units (GPU). Pure Data, thanks to its graphical modular development environment, allows fast prototype developments. Those factors led us to start a research program dedicated to the realizations of image processing modules for Pure Data written in CUDA. So in this we have implemented image blurring algorithm to get familiar with GPGPU.

Scientific Cloud Computing has emerged rapidly in the last few years. Scientific Cloud Computing, which targets high performance communities, has caused a shift over of the scientific applications running on conventional physical infrastructure to a more scalable and flexible virtual infrastructure managed by the cloud providers. Scientific Cloud Computing has brought up a lot of advantages (pay-per-use, resource on-demand, high-availability, flexibility, scalability) over the traditional computing models. Traditional computing models, in order to use scientific applications, require in-house infrastructure (servers, storage etc.) and skill to manage this infrastructure, whereas the Scientific Cloud Computing allows the user to focus on applications without worrying about the infrastructure and installations required. Scientific clouds need proper access mechanisms to improve the adoption of such facilities by scientific communities and enhance the usage of scientific clouds. It is very important to have a proper access mechanism in place for Scientific Clouds so as to improve the adoption of such facilities and enhance the usage of cloud. In this paper, the authors present a Web-Based Job Submission Mechanism for Scientific Cloud Computing, SciInterface. SciInterface provides a graphical user-friendly interface to the users for interacting with the Scientific Cloud and a facility to submit scientific jobs in to the cloud for processing. The prominent features of SciInterface are user privileges mapping, job submission, job monitoring and output logs. The authors, in this paper, discuss the architecture and features of SciInterface, along with the challenges faced during the development.

Steering and guiding an endoscope in a patient's body has been a challenge for doctors for many years now. Present endoscopic technology allows passive bending of the scope by one or more cables attached to the tip. Bending motions of the tip of a conventional endoscope are controlled from outside the body by wire traction. Therefore, precise operation of the endoscope is difficult in areas of complex shape such as the intestine. Furthermore, patients suffer pain during a procedure with an endoscope. This paper aims at deciphering a solution for the above problems by introducing the concept of active bending of the endoscope with Shape Memory Alloy (SMA) micro-actuators. A Teflon prototype is developed to establish the concept. Many Teflon segments are connected with ball and socket joint and each segment bending is controlled independently which gives high degree of maneuverability to the device. The dexterous device prototype developed is capable of following complex trajectories and can steer tight bends. The same concept of bending can be extended to devices like borescopes, to be used for internal visualization of mechanical systems such as jet engines, turbines, satellites etc. These devices are again expected to follow complex trajectories with tight bends for visualization of specific areas within the highly integrated structure without dismantling.

Parallel coupled microstrip filters are extensively used as band pass filter in microwave communication because they are compact in size, easy to fabricate due to its planar structure. The responses of these filters are satisfactory for narrow bandwidth due to their relatively weak coupling. A severe limitation of the conventionally designed parallel coupled filters is parasitic response at harmonics. In this paper some techniques are discussed to minimize the spurious bands, and also an effective design to suppress these spurious pass bands has been proposed.

The fractal geometries such as Koch curve, Sierpinsky gasket, and Hilbert curve, etc. have been widely studied and applied to design several microwave devices, such as antennas, frequency selective surfaces (FSS), and photonic band gap (PBG) devices. There are several advantages of these fractal devices including reduction of resonant frequencies, smaller size and broadband width. In this paper a critical study has been presented on application of fractal geometry to design the microstrip antenna, microstrip filters and frequency selective surfaces.