Processing Power Requirements Research Articles

Audio-based event detection in the operating room.

Even though workflow analysis in the operating room has come a long way, current systems are still limited to research. In the quest for a robust, universal setup, hardly any attention has been given to the dimension of audio despite its numerous advantages, such as low costs, location, and sight independence, or little required processing power. We present an approach for audio-based event detection that solely relies on two microphones capturing the sound in the operating room. Therefore, a new data set was created with over 63h of audio recorded and annotated at the University Hospital rechts der Isar. Sound files were labeled, preprocessed, augmented, and subsequently converted to log-mel-spectrograms that served as a visual input for an event classification using pretrained convolutional neural networks. Comparing multiple architectures, we were able to show that even lightweight models, such as MobileNet, can already provide promising results. Data augmentation additionally improved the classification of 11 defined classes, including inter alia different types of coagulation, operating table movements as well as an idle class. With the newly created audio data set, an overall accuracy of 90%, a precision of 91% and a F1-score of 91% were achieved, demonstrating the feasibility of an audio-based event recognition in the operating room. With this first proof of concept, we demonstrated that audio events can serve as a meaningful source of information that goes beyond spoken language and can easily be integrated into future workflow recognition pipelines using computational inexpensive architectures.

GPU optimization techniques to accelerate optiGAN—a particle simulation GAN

The demand for specialized hardware to train AI models has increased in tandem with the increase in the model complexity over the recent years. Graphics processing unit (GPU) is one such hardware that is capable of parallelizing operations performed on a large chunk of data. Companies like Nvidia, AMD, and Google have been constantly scaling-up the hardware performance as fast as they can. Nevertheless, there is still a gap between the required processing power and processing capacity of the hardware. To increase the hardware utilization, the software has to be optimized too. In this paper, we present some general GPU optimization techniques we used to efficiently train the optiGAN model, a Generative Adversarial Network that is capable of generating multidimensional probability distributions of optical photons at the photodetector face in radiation detectors, on an 8GB Nvidia Quadro RTX 4000 GPU. We analyze and compare the performances of all the optimizations based on the execution time and the memory consumed using the Nvidia Nsight Systems profiler tool. The optimizations gave approximately a 4.5x increase in the runtime performance when compared to a naive training on the GPU, without compromising the model performance. Finally we discuss optiGANs future work and how we are planning to scale the model on GPUs.

Lightweight and Robust Malware Detection Using Dictionaries of API Calls

Malware in today’s business world has become a powerful tool used by cyber attackers. It has become more advanced, spreading quickly and causing significant harm. Modern malware is particularly dangerous because it can go undetected, making it difficult to investigate and stop in real time. For businesses, it is vital to ensure that the computer systems are free from malware. To effectively address this problem, the most responsive solution is to operate in real time at the system’s edge. Although machine learning and deep learning have given promising performance for malware detection, the significant challenge is the required processing power and resources for implementation at the system’s edge. Therefore, it is important to prioritize a lightweight approach at the system’s edge. Equally important, the robustness of the model against the concept drift at the system’s edge is crucial to detecting the evolved zero-day malware attacks. Application programming interface (API) calls emerge as the most promising candidate to provide such a solution. However, it is quite challenging to create API call features to achieve a lightweight implementation, high malware detection rate, robustness, and fast execution. This study seeks to investigate and analyze the reuse rate of API calls in both malware and goodware, shedding light on the limitations of API call dictionaries for each class using different datasets. By leveraging these dictionaries, a statistical classifier (STC) is introduced to detect malware samples. Furthermore, the study delves into the investigation of model drift in the STC model, employing entirely distinct datasets for training and testing purposes. The results show the outstanding performance of the STC model in accurately detecting malware, achieving a recall value of one, and exhibiting robustness against model drift. Furthermore, the proposed STC model shows comparable performance to deep learning algorithms, which makes it a strong competitor for performing real-time inference on edge devices.

Targeted sequencing analysis pipeline for species identification of human pathogenic fungi using long-read nanopore sequencing

Among molecular-based techniques for fungal identification, Sanger sequencing of the primary universal fungal DNA barcode, the internal transcribed spacer (ITS) region (ITS1, 5.8S, ITS2), is commonly used in clinical routine laboratories due to its simplicity, universality, efficacy, and affordability for fungal species identification. However, Sanger sequencing fails to identify mixed ITS sequences in the case of mixed infections. To overcome this limitation, different high-throughput sequencing technologies have been explored. The nanopore-based technology is now one of the most promising long-read sequencing technologies on the market as it has the potential to sequence the full-length ITS region in a single read. In this study, we established a workflow for species identification using the sequences of the entire ITS region generated by nanopore sequencing of both pure yeast isolates and mocked mixed species reads generated with different scenarios. The species used in this study included Candida albicans (n = 2), Candida tropicalis (n = 1), Nakaseomyces glabratus (formerly Candida glabrata) (n = 1), Trichosporon asahii (n = 2), Pichia kudriavzevii (formerly Candida krusei) (n = 1), and Cryptococcus neoformans (n = 1). Comparing various methods to generate the consensus sequence for fungal species identification, the results from this study indicate that read clustering using a modified version of the NanoCLUST pipeline is more sensitive than Canu or VSEARCH, as it classified species accurately with a lower abundance cluster of reads (3% abundance compared to 10% with VSEARCH). The modified NanoCLUST also reduced the number of classified clusters compared to VSEARCH, making the subsequent BLAST+ analysis faster. Subsampling of the datasets, which reduces the size of the datasets by approximately tenfold, did not significantly affect the identification results in terms of the identified species name, percent identity, query coverage, percentage of reads in the classified cluster, and the number of clusters. The ability of the method to distinguish mixed species within sub-populations of large datasets has the potential to aid computer analysis by reducing the required processing power. The herein presented new sequence analysis pipeline will facilitate better interpretation of fungal sequence data for species identification.

Overview of the Low Complexity Enhancement Video Coding (LCEVC) Standard

The Low Complexity Enhancement Video Coding (LCEVC) specification is a recent standard approved by the ISO/IEC JTC 1/SC 29/WG04 (MPEG) Video Coding. The main goal of LCEVC is to provide a standalone toolset for the enhancement of any other existing codec. It works on top of other coding schemes, resulting in a multi-layer video coding technology, but unlike existing scalable video codecs, adds enhancement layers completely independent from the base video. The LCEVC technology takes as input the decoded video at lower resolution and adds up to two enhancement sub-layers of residuals encoded with specialized low-complexity coding tools, such as simple temporal prediction, frequency transform, quantization, and entropy encoding. This paper provides an overview of the main features of the LCEVC standard: high compression efficiency, low complexity, minimized requirements of memory and processing power.

A comparitive analysis of mobilenet and xception architecture for classification of endoscopy images

The deep architectures have been making their significant mark in solving the image classification problems. The usage of the standard deep learning architectures has been increasing at a greater pace because of the outcome of the state-of-the art accuracy and precision by incorporating these architectures in image classification problems. The classification of medical images from variety of modalities is also possible with better accuracy by using these deep architectures. The standard architectures that have achieved substantial results in general object classification can be used and their learning can be transferred to the required domain. Further, these architectures can be used in an ensemble structure thereby creating a customized architecture for solving the classification problem. Some of the challenges in including these architectures include, the enormous memory and processing power requirement. Current trends are focused towards generating optimized deep structures that has relatively lesser memory and processing capacity requirement. To this end, the two benchmark architectures that have obtained momentous results in object classification namely the MobileNetV2 and Xception are included in this study to perform binary classification of capsule endoscopy images to detect the presence of abnormality. Capsule endoscopy (CE) is a medical procedure where the patient swallows the pill-based camera that traverses the entire gastrointestinal (GI) path capturing numerous images. These images are to be assessed for detecting the presence of any abnormality. The architectures included in this work have been trained on the images from the publicly available CE datasets and an ablation study on the hyperparameter tuning of the architecture have been conducted. The results are compared, analyzed and presented here. The MobileNetV2 and Xception architecture have achieved a maximum accuracy of 85% and 82% in the abnormality detection in CE images.

Requirements and Hardware Limitations of High-Frame-Rate 3-D Ultrasound Imaging Systems

The spread of high frame rate and 3-D imaging techniques has raised pressing requirements for ultrasound systems. In particular, the processing power and data transfer rate requirements may be so demanding to hinder the real-time (RT) implementation of such techniques. This paper first analyzes the general requirements involved in RT ultrasound systems. Then, it identifies the main bottlenecks in the receiving section of a specific RT scanner, the ULA-OP 256, which is one of the most powerful available open scanners and may therefore be assumed as a reference. This case study has evidenced that the “star” topology, used to digitally interconnect the system’s boards, may easily saturate the data transfer bandwidth, thus impacting the achievable frame/volume rates in RT. The architecture of the digital scanner was exploited to tackle the bottlenecks by enabling a new “ring“ communication topology. Experimental 2-D and 3-D high-frame-rate imaging tests were conducted to evaluate the frame rates achievable with both interconnection modalities. It is shown that the ring topology enables up to 4400 frames/s and 510 volumes/s, with mean increments of +230% (up to +620%) compared to the star topology.

Trust-aware and incentive-based offloading scheme for secure multi-party computation in Internet of Things

Adoption of multi-party computation in IoT provides the required processing power needed by the IoT devices to provide smart services in the shortest time. However, this requires a secure offloading scheme that is capable of fairly offloading the computations of source nodes to a different set of workers, guarantees the privacy of the source nodes, and verifies the correctness of results without a third party at a low overhead.In this paper, we formulate a secure privacy-preserving offloading scheme based on modified secret sharing to offload computations and data to a different set of workers. We also develop incentive and trust models to encourage honesty and willingness and discourage delay among workers during multi-party computation. Last, we develop a low overhead morphism-based verification technique for the self-verification of the correctness of results. We finally present the security analysis of the scheme which shows that the schemes meet up with the necessary security requirements, and the experimental results show the capability of the scheme in terms of its security functionalities, low computation cost, effective verification of results, and generation of incentives and trust values for workers during multi-party computation.

Increasing privacy and security by integrating a Blockchain Secure Interface into an IoT Device Security Gateway Architecture

Internet of Things and Blockchain are considered two major technologies. Lower latency and a higher linked system number provide greater flexibility for remote execution of Internet of Things (IoT) applications. It is no secret that IoT devices often have insufficient computing capacity (both in terms of processing power and storage requirements) to support robust protection and encryption algorithms. The Internet of Things is facing many challenges such as poor interoperability, security vulnerabilities, privacy, and lack of industry standards. Cyber-attacks on IoT devices can have an impact on energy trading privacy and security. This paper suggests a method for introducing a basic interface to an IoT device’s security gateway architecture along with Blockchain to provide decentralization and authentication. It adds much-needed anonymity and versatility to IoT infrastructure, which is currently lacking. The solution enhances the reliability of data sent to remote services by applying compatible cryptographic algorithms to it before sending it. The solution’s benefits include compatibility with all IoT products and the ability to run any cryptographic algorithm on data that can be used for microgrid trading and can be initialized and securely transported over 5G or 6G network infrastructures. As a part of this work, a security procedure has been created that supports every cryptographic algorithm for all IoT devices in the network. In addition, the interface is guarded by the Blockchain technology which eliminates single control authority, records historical transactions performed by the IoT devices and provides a trust between devices.

A smartphone-based multimodal indoor tracking system

The great popularity of smartphones, together with the increasingly important aim of providing context-aware services, has spurred interest in developing indoor tracking systems. Accurate tracking and localization systems are seen as key services for most context-aware applications. Research projects making use of radio signals detected by radio interfaces and the data captured by sensors commonly integrated in most smartphones have already shown promising and better results than location solutions based on a single data source. In this paper, we present a multi-sensor tracking system built by incrementally integrating state-of-the-art models of the Wi-Fi interface and the accelerometer, gyroscope and magnetometer sensors of a smartphone. Our proposal consists of a simple calibration phase of the tracking system, which involves enabling simultaneous data gathering from all three sensors and the Wi-Fi interface. Taking the Wi-Fi signal model as baseline, four different configurations are evaluated by incrementally adding and integrating the models of the other three sensors. The experimental results reveal a mean error accuracy of 60 cm in the case when the tracking system makes use of all four data sources. Our results also include a spatial characterization of the accuracy and processing power requirements of the proposed solution. Our main findings demonstrate the feasibility of developing accurate localization indoor tracking systems using current smartphones without the need for additional hardware.

Derin Öğrenme İle Yaprak Sınıflandırmada Görüntü Boyutu, Arka Plan Rengi ve Gri Resim ile Renkli Resim Arasındaki Farkın İncelenmesi

In academic studies, there are many factors that change depending on the changes in the parameters of the process, such as the processing time, the required processing power, as well as the success. In the methods used for classification, recognition, and detection, the changes in the data received as input may affect the result, as well as the variables specific to the methods used. Convolutional neural networks, whose use is increasing day by day in processes such as classification and recognition using images, learn and classify the characteristics of data sets in different image sizes, including color, gray or black and white images, with filters and functions on the layers in the model. Many different parameters such as layers in the created model and filters and functions in these layers can be changed. As a result of these changes, the most suitable number of layers, the optimum values for the parameters and functions in these layers are determined for the data set used. There are studies focused on optimizing many different structures, such as reproducing the images in the used data set or determining the best by testing different parameters in the classification method. In this study, while the changes were made in the leaf images with a fixed background in the determined leaf data set, the model used in leaf classification with convolutional neural network was kept constant. It is aimed to examine the pictures used for 3 different image sizes, the gray picture or color picture difference and the changes caused by the background color.

Solution-processed oxide semiconductor-based artificial optoelectronic synapse array for spatiotemporal synaptic integration

Optoelectronic neuromorphic systems mimicking the structure and the signal processing of biological neural systems have gained significant interest due to their potential advantages such as high computing speed, high bandwidth, parallel processing, and low power requirements. Here, we demonstrate an optoelectronic synapse array consisting of solution-processed indium-gallium-zinc-oxide (IGZO) synapse devices emulating complex neural functions such as the spatiotemporal synaptic integration for neuromorphic implementation. Particularly, we adopted a vertically stacked metal-insulator-semiconductor-metal structure for IGZO synapse device to enable efficient light-induced conductance update and a low-voltage operation typically below 3 V. Using light as the stimulation, versatile synaptic functions including short-term memory/long-term memory, symmetric spike-timing dependent plasticity, and spike-number dependent plasticity were mimicked. In addition, we obtained a high recognition accuracy of handwritten digit images up to 89.4% by implementing ADAM training algorithm. Furthermore, using a crossbar structure 3 × 3 IGZO synapse array, the emulation of spatiotemporal summation was successfully carried out which is believed to be responsible for the neural encoding and the auditory recognition processes occur in the brain.

Continuous Lyapunov Controlled Non-linear System Optimization Using Deep Learning with Memory

The introduction of unexpected system disturbances and new system dynamics does not allow guaranteed continuous system stability. In this research we present a novel approach for detecting early failure indicators of non-linear highly chaotic system and accordingly predict the best parameter calibrations to offset such instability using deep machine learning regression model. The approach proposed continuously monitors the system and controller signals. The Re-calibration of the system and controller parameters is triggered according to a set of conditions designed to maintain system stability without compromise to the system speed, intended outcome or required processing power. The deep neural model predicts the parameter values that would best counteract the expected system in-stability. To demonstrate the effectiveness of the proposed approach, it is applied to the non-linear complex combination of Duffing Van der pol oscillators. The approach is also tested under different scenarios the system and controller parameters are initially chosen incorrectly or the system parameters are changed while running or new system dynamics are introduced while running to measure effectiveness and reaction time.

Matrix multiplication by neuromorphic computing

With increased requirements of processing power in many computing applications, and the consequent huge increases in resource costs including electrical power consumption, there is a push to find alternative computing architectures that achieve high speeds but require less resources. One direction that shows promise is brain-inspired neuromorphic computing; it has thus become an active area of research. Much of the current work in neuromorphic computing targets its applicability in solving machine learning and pattern recognition problems. We push this boundary to show the possibility of using neuromorphic architectures to solve matrix multiplication, a core mathematical task pertinent to a host of engineering problems. We use a neuromorphic simulator to present a solution to matrix multiplication that demonstrates a significant performance improvement, both in terms of processing time and intermediate storage over existing approaches. The basic principle of proposed method is operating on a single intermediate transformation matrix that is flattened into an ensemble array represented as a neural node that is operated on. This is an optimisation over memory and time over the existing approach which uses two sparse intermediate matrices. Theoretical and experimental analysis indicates the importance of the proper choice of neuromorphic parameters, specifically the number of neurons and radius of the ensemble along with the dimensions and elements in input matrices has on accuracy, time and memory. We indicate an approach to compute the proper choice of neuromorphic resources given relevant benchmark constraints.

AgriTrust-A Trust Management Approach for Smart Agriculture in Cloud-based Internet of Agriculture Things.

Internet of Things (IoT) provides a diverse platform to automate things where smart agriculture is one of the most promising concepts in the field of Internet of Agriculture Things (IoAT). Due to the requirements of more processing power for computations and predictions, the concept of Cloud-based smart agriculture is proposed for autonomic systems. This is where digital innovation and technology helps to improve the quality of life in the area of urbanization expansion. For the integration of cloud in smart agriculture, the system is shown to have security and privacy challenges, and most significantly, the identification of malicious and compromised nodes along with a secure transmission of information between sensors, cloud, and base station (BS). The identification of malicious and compromised node among soil sensors communicating with the BS is a notable challenge in the BS to cloud communications. The trust management mechanism is proposed as one of the solutions providing a lightweight approach to identify these nodes. In this article, we have proposed a novel trust management mechanism to identify malicious and compromised nodes by utilizing trust parameters. The trust mechanism is an event-driven process that computes trust based on the pre-defined time interval and utilizes the previous trust degree to develop an absolute trust degree. The system also maintains the trust degree of a BS and cloud service providers using distinct approaches. We have also performed extensive simulations to evaluate the performance of the proposed mechanism against several potential attacks. In addition, this research helps to create friendlier environments and efficient agricultural productions for the migration of people to the cities.

An asymmetrical multilevel inverter with optimum number of components based on new basic structure for photovoltaic renewable energy system

In this paper, a new basic structure of asymmetrical multilevel inverter with optimum number of components for photovoltaic (PV) renewable energy system is proposed. The aim is to reduce the number of components required to achieve a high number of output levels compared to older topologies. The inverter configuration comprises of two basic units connected to a packed H-Bridge. Cascaded operation of the inverter configuration is feasible for higher level generation. The magnitudes of the DC sources are selected in binary sequence to maximize the number of output levels. Solar panels with proportional-integral (PI) controlled DC-DC converters are used as the DC sources. The switch modulations are achieved by pre-defined switching angles which are calculated to minimize total harmonic distortion (THD) in the AC output waveforms. This method is used instead of the classic pulse width modulation (PWM) technique to reduce the total switching loss and the required processing power in generating the switching signals. Compared to the conventional multilevel inverter topologies, the proposed inverter is a better choice in terms of the number of components. It also achieves a good balance between the number of components and total blocking voltage. Hence, its installation will require smaller space and cost. Another merit of the proposed topology is that it produces AC output with low THD at high efficiency. The efficacy of the proposed asymmetrical 31-level inverter topology is simulated using MATLAB/Simulink and is further verified through experimental analysis.

An enhanced decentralized artificial immune-based strategy formulation algorithm for swarms of autonomous vehicles

This work presents an algorithmic approach to the problem of strategy assignment to the members of a swarm of autonomous vehicles. The proposed methodology draws inspiration from the artificial immune system (AIS), where a large number of antibodies cooperate in order to protect an organism from foreign threats by local exchange of information. The decentralized nature of the methodology does not suffer from problems like the need of a central control unit, the high maintenance costs and the risks associated with having a single point of system failure, which are common to centralized control techniques. Decentralized and distributed optimization schemes employ simple algorithms, which are fast, robust and can run locally on an autonomous unit due to their low processing power requirements. In contrast to standard AIS-based decentralized schemes, the proposed methodology makes use of a dynamic formulation of the available strategies and avoids the possibility of choosing an invalid strategy, which may lead to inferior swarm performance. The methodology is further enhanced by a dual strategy activation decay technique and a blind threat-follow rule. Statistical testing on different case studies based on “enemy search and engage” type scenarios in a simulated environment demonstrates the superior performance of the proposed algorithm against the standard AIS, an enhanced AIS version and a centralized particle swarm optimization (PSO) based methodology.

Development and Innovation of Extractor For Malunggay (Moringa oleifera Lamk)

This study aimed to develop and innovate an extractor for malunggay to solve its laborious and tedious processing and for product and by-products development. Particularly, innovation of existing malunggay juice extractor was done and tested in terms of efficiency, capacity, extraction time, losses and power requirement, and its technical and economic viability. The extractor was innovated for extracting malunggay juice and other similar commodities. Testing of innovated extractor was highly significant in terms of extraction time, volume of extracted juice, and percent recovery of juice extracted compared to the manual extraction process. The juice extraction of the malunggay using the innovated extractor showed a capacity of 5.5 kg/hr to almost 19 kg/hr using matured and young leaves and stalks. The mean juice recovery is 62%. The innovated extractor for malunggay has higher efficiency and capacity but has lesser processing time, losses, and power requirement. The malunggay extractor is economically viable and feasible. The extractor could extract juice from malunggay, and its meal could be converted into powder and tea and gain a minimum of ₱62/kg of malunggay. It has a return on investment of 34.11% with a payback period of 2.1 years. The fabrication of the prototype of “Agiseparator,” downsizing of the equipment, technology commercialization, adoption of a backyard planting for malunggay; health impact dissemination, determination of nutrient retention and loss during processing, and development of more products from malunggay are recommended.

Distributed clustering and operational state scheduling in wireless rechargeable sensor networks

Replacement of exhausted batteries in wireless sensor networks might lead to temporal disruption of network operations. Wireless rechargeable sensor networks (WRSNs) can mitigate this problem by recharging nodes before they run out of batteries. However, because WSRN nodes cannot perform recharging and task monitoring simultaneously, it is challenging to ensure continuous network operations while maintaining low-cost and low processing-power requirements of nodes. It is necessary to design an effective and energy-efficient node scheduling scheme to schedule nodes to either monitoring or recharging state without adversely reducing the network lifetime and throughput. In this paper, we propose a distributed clustering and operational state scheduling algorithm (DCOS) to maximise the overall network lifetime. Nodes needing energy can be replenished within time intervals when they are not in the state of sensing and transmitting. We conducted extensive simulations to evaluate the performance of DCOS. The results show that DCOS outperforms most of the state-of-the-art methods.

An IoT-based Smart Orchard Monitoring System by Employing Wireless Sensor Networks

In this study, an IoT-based smart orchard monitoring is proposed to gather and transmit environment data from a sensor node to a central node for necessary and relevant actuation in order to have good produce at the soonest amount of time. Wireless sensor motes are deployed based on a simple linear pattern across a square farm and only require the minimum set of specifications to monitor its surrounding. On the other hand, the central nodes will require more processing power, memory and power requirements. Sensor and central nodes communicate in a line-of-sight method and follows a deterministic routing table based on the sensor node’s four neighbors. Throughput, latency, and energy consumption results are presented to allow designers and farmers consideration and freedom on how to select which routing protocol can be used to achieve their target objectives.