This paper presents a step-up DC-DC converter that uses a stepwise gate-drive technique to reduce the power FET gate-drive energy by 82%, allowing positive efficiency down to an input voltage of ±0.5 mV—the lowest input voltage ever achieved for a DC-DC converter as far as we know. Below ±0.5 mV the converter automatically hibernates, reducing quiescent power consumption to just 255 pW. The converter has an efficiency of 63% at ±1 mV and 84% at ±6 mV. The input impedance is programmable from 1 Ω to 600 Ω to achieve maximum power extraction. A novel delay line circuit controls the stepwise gate-drive timing, programmable input impedance, and hibernation behavior. Bipolar input voltage is supported by using a flyback converter topology with two secondary windings. A generated power good signal enables the load when the output voltage has charged above 2.7 V and disables when the output voltage has discharged below 2.5 V. The DC-DC converter was used in a thermoelectric energy harvesting system that effectively harvests energy from small indoor temperature fluctuations of less than 1°C. Also, an analytical model with unprecedented accuracy of the stepwise gate-drive energy is presented.

Neurons can code for multiple variables simultaneously and neuroscientists are often interested in classifying neurons based on their receptive field properties. Statistical models provide powerful tools for determining the factors influencing neural spiking activity and classifying individual neurons. However, as neural recording technologies have advanced to produce simultaneous spiking data from massive populations, classical statistical methods often lack the computational efficiency required to handle such data. Machine learning (ML) approaches are known for enabling efficient large scale data analyses; however, they typically require massive training sets with balanced data, along with accurate labels to fit well. Additionally, model assessment and interpretation are often more challenging for ML than for classical statistical methods. To address these challenges, we develop an integrated framework, combining statistical modeling and machine learning approaches to identify the coding properties of neurons from large populations. In order to demonstrate this framework, we apply these methods to data from a population of neurons recorded from rat hippocampus to characterize the distribution of spatial receptive fields in this region.

Over the past decade, neural networks have shown commendable performance in complex tasks such as Image Recognition, Natural Language Processing, and game playing. At the same time, these networks have unfortunately been proven to be susceptible to minor changes to their inputs, commonly called adversarial perturbations. For example, an otherwise accurate DNN model can be deceived into misclassifying a given image by making slight alterations to the image, invisible to human eyes. Such adversarial attacks make it challenging to deploy them in security-sensitive applications such as autonomous vehicles. It is, therefore, crucial to understand adversarial attacks to test the robustness of a model and come up with suitable defense techniques. This paper aims to survey popular works in this domain. In our discussion, we broadly classify the attacks as image-dependent or independent and targeted or untargeted. We also cover some interesting defense techniques. We discuss real life or physical attacks where an attacker can perturb the image in real life, for instance, by painting a stop sign or pasting stickers on objects. We also show some experimental results, as we test the performance of two adversarial attacks against a pre-trained VGG19 model on Imagenet dataset.

Sustainability and climate monitoring efforts create a need for long-term remote sensing of  large geographic areas.  However, environmental monitoring in remote areas of developing countries remains impeded by a lack of low-cost, scalable IoT~solutions.  Whereas IoT systems for remote sensing abound,  they mostly are either low-cost or suitable for large areas, but not both. In this paper, we present a low-cost low-power network solution for remote sensing of areas up to hundreds of square kilometers.  Taking advantage of LoRa technology, we develop a self-organizing mesh network that can be scaled to a hundred and more nodes. Scalability is achieved by developing methods that mitigate packet collisions during data collection.  We present a protocol, called CottonCandy, with which nodes self-organize in a spanning-tree network topology in a distributed fashion. A power profile on a  custom-built circuit board shows that CottonCandy nodes can run  thousands of duty cycles on 2~AA batteries, sufficient to operate for years in many applications. Using off-the-shelf components, the cost of a CottonCandy node is less than US-$ 15.  Evaluations by simulation show that CottonCandy networks with 100 nodes achieve a packet delivery ratio of >90%. Measurements of an outdoor deployment with 15~nodes corroborate the high packet delivery ratio in a real-life setting.

A low mutually coupled multiple-input multiple-output (MIMO) antenna is designed for on-chip millimetre-wave (mm-Wave) application. Two circularly polarized (CP) cylindrical dielectric resonator antennas (CDRA) with polarization diversity are placed over a double substrate silicon layer. Herein, a simple decoupling approach is applied to inhibit mutual coupling by introducing a metal plate in the substrate between the two DRAs. According to the simulation results, the inserted metallic plate improved isolation by 4-5 dB in the desired CP bandwidth. This is achieved while maintaining excellent radiation efficiency and gain, making the suggested antenna appropriate for mm-wave on-chip MIMO applications. (This paper has been submitted to  Asia-Pacific Microwave Conference 2022 (APMC 2022) )

In this paper, we study the personalized ranking problem for item’s recommendation text (product description). Item’s recommendation text is the sentence that describe the item highlights for user decision (such as buy or click). The recommendation texts is shown under the item, and we also call them rec-texts. One item has multiple rec-texts, and different rec-texts have different affect for user decision. So the problem is to capture the user preference for each item’s rec-texts and to personalized display. In this paper, we study multiple methods to train a model which learn the user preference scores for each item’s rec-texts. The online experimental results demonstrate our methods work.

With a mockup prototype of a robotic hand, various algorithms were implemented. This exposed a broad range of applications of the prototype. In this body of work, algorithms of a dexterous mechanical robotic hand will be presented which span areas including recreation, healthcare and education. Imagine a child learning to count. If this hypothetical child had a robotic assistant who could infinitely demonstrate the way to count without getting tired, consequently the child would learn how to count at a much-accelerated rate. Perhaps in an arcade there is a robotic hand that plays games like rock paper scissors or thumb wars, subsequently the youth will be further inspired to pursue an understanding of robotic systems. At the Tech Crunch Robotic conference on July 21st, Dean Kamen the inventor of the segway said something very profound, it was along the lines of how when it comes to interest in Robotics, we are competing with activities like after school sports or video games because these are the things that consume the time of children. However, robotics is mostly seen as an academic activity when it must not be so. With the changing job market, equipping arcade games with intelligent robots is not a farfetched idea. Another algorithm implemented was a sign language algorithm where the robotic hand executes a sequence of hand-signs in a desired order. The programming was done with Adafruit's PCA9685 controller which can control 16 servo motors through an I2C interface. When all I2C buses are used then the controller can control up to 992 servo motors.

This paper deals with peak to average power (PAPR) reduction in a single and multi-user orthogonal chirp division multiplexing (OCDM) context. Two methods for PAPR reduction based on the selection of the frequency variation (up or down) of the chirps are first presented in a single user system. The first technique consists in considering two OCDM signals generated with up and down chirps, respectively, and selecting the one offering lowest PAPR. The second PAPR reduction method is based on usual clipping, and in that case the chirp selection aims to reduce the clipping noise. An adapted receiver is also presented, based on the maximum likelihood (ML) estimation of the frequency variation (up or down) of the chirp. Then, a general procedure for multi-user OCDM transmission is introduced, where a sub-band of the available bandwidth is dedicated to each user, whose frequency of the chirps varies within this sub-band. Next, the PAPR reduction techniques are generalized to this multi-user OCDM system. Moreover, a performance analysis of the first PAPR reduction method is developed, and supported by simulation results. It is also shown that the chirp selection reduces the clipping noise, and improves the bit error rate performance compared with clipping only.

We propose an exact new sampling algorithm using classical statistical theory. The algorithm is powered by white noise and non-local information. The proposed method will process an infinite number of rejected samples in parallel before accepting one exact single sample from a target distribution using an optimization mechanism. The algorithm is applied to sampling from general distributions in addition to sampling from the wave function density distribution which lies at the heart of the measurement problem in quantum physics.

The tunable electrical properties of conducting polymers (CPs), biocompatibility, fabrication versatility, and cost-efficiency make them an ideal coating material for stimulation electrodes in biomedical applications. Several biological processes like wound healing, neuronal regrowth, and cancer metastasis, which rely on constant electric fields, demand electrodes capable of delivering direct current stimulation (DCs) for long times without developing toxic electrochemical reactions. Recently, CPs such as poly(3,4-ethylenedioxythiophene) polystyrene sulfonate (PEDOT/PSS) have demonstrated outstanding capability for delivering DCs without damaging cells in culture while not requiring intermediate buffers, contrary to the current research setups relying on noble-metals and buffering bridges. However, a clear understanding of how electrode design and CP synthesis influence DCs properties of these materials is still needed. This study demonstrates that various PEDOT-based CP coatings and hydrogels on rough electrodes can deliver DCs without substantial changes to the electrode and noticeable development of chemical byproducts depending on the electrode area and polymer thickness. A comprehensive analysis of the tested coatings is provided according to the desired application and available resources alongside a proposed explanation for the observed electrochemical behavior. The CPs tested herein can pave the way toward the widespread implementation of DCs as a therapeutic stimulation paradigm.

A study of the phase synchronization transient of three identical linear LC oscillators, coupled through identical resistors is presented. Exact equations that represent, as a function of time, the evolution of the phase angles between the oscillators, from the initial condition until the synchronization are provided.

The method of Kobayashi potentials (KP) in Electromagnetics (EM) is presented in a simple manner. The target audience is graduate students. Diffraction by a flanged parallel-plate waveguide is selected as a sample problem.

In this paper, the diversity order (DO) and the diversity measure (DM) of multiple-input multiple-output (MIMO) antennas in generic single-user, multi-user and massive MIMO wireless communications are calculated for several practical case studies. These performance metrics are calculated from the antenna correlation coefficient (ACC) matrix. The DO is calculated from the rank of the ACC matrix, and the DM is obtained from the trace and Frobenius norm of the ACC matrix. Numerical results are provided. In particular, it is shown that though the DM is always less than the DO, it incorporates the effect of antenna correlation on the diversity gain in a better way.  (This paper has been submitted to  IEEE Antennas and Wireless Propagation Letters )

This work studies the potential of transforming STEM education with the integration of recent STEM learning technologies. STEM is the focus of the education system as it is the driver for creation, innovation, and invention. However, the delay in the introduction of STEM in the education system impacts the motivation of youths to take up STEM courses, thereby adversely affecting the enrolment into the field and proliferation. Furthermore, the digital divide among the youths in urban and rural settings also plays a vital role in creating the gap when it comes to STEM education. Therefore, this work analyses the experimental-based case study to assess the impact of early access to STEM educational tools in ensuring the youths understand the potential of STEM. Priliminary study to understand the pupil’s perception on existing method of STEM teaching-learning is carried out. The general perception is that students’ are not able to visualize, neither they are able to understand the potential of STEM. Hence, does not play significant role in motivating the pupil to take up the STEM fields. On the other hand, as part of the study, same group who responded in preliminary study took part in STEM workshops. Basic concepts and applications of automation, the internet of things, and AI were considered for the students to visualize the potential of STEM education. STEMSEL microncontroller and runlinc- an online IoT and AI application development platform were used as study tool. Students demonstrated their ability to discover their own creativity and innovation using the platform where 3D model (design, development, and deployment) concept was inculcated. The participant students were from schools in rural location who do have access to such learning platforms. On completion of workshops, pupils’ (students) understanding of STEM, their perception towards it, and their visualization on the potential of STEM education in making the impact for social good drastically improved. Therefore, the study indicated that the early access to hands-on educational tools, integration into teaching-learning of STEM would bring transformation in STEM education. Such exposure would also motivate youths take up STEM fields in higher education, thereby enhancing the enrollment.

Progress in automatic control of thermal processes and real-time estimation of heat penetration into live tissue has long been limited by the difficulty of obtaining high-fidelity thermodynamic models. Traditionally, in complex thermodynamic systems, it is often infeasible to estimate the thermophysical parameters of spatiotemporally varying processes, forcing the adoption of model-free control architectures. This comes at the cost of losing any robustness guarantees, and implies a need for extensive real-life testing. In recent years, however, infrared cameras and other thermographic equipment have become readily applicable to these processes, allowing for a real-time, non-invasive means of sensing the thermal state of a process. In this work, we present a novel physics-based approach to learning a thermal process’s dynamics directly from such real-time thermographic data, while focusing attention on regions with high thermal activity. We call this process, which applies to any higher-dimensional scalar field, attention-based noise robust averaging (ANRA). Given a partial-differential equation model structure, we show that our approach is robust against noise, and can be used to initialize optimization routines to further refine parameter estimates. We demonstrate our method on several simulation examples, as well as by applying it to electrosurgical thermal response data on in vivo porcine skin tissue.

Freezing of gait (FoG) is a widely observed movement disorder in Parkinson’s Disease patients (PD). Its prediction is crucial for effectively giving the cue to avoid FoG occurrence. However, present methods of prediction of FoG are inaccurate for large but practical prediction horizons (PH)s. Therefore, this work presents a comprehensive analysis of the electroencephalography (EEG) and inertial measurement units (IMU)s to predict FoG advance in time. An ensemble model consisting of two neural networks, EEGFoGNet and IMUFoGNet, was developed and tested at different PHs and ensemble weights. Moreover, the model is tested for two practical scenarios: clinical or research applications and personal uses. For clinical or research applications, stratified 5-Folds cross-validation was used. For personal uses, a transfer learning technique was used for learning user-specific FoG-related features. The model obtained the best accuracy of 92.1% at 1 second’s PH and the least accuracy of 86.2% at 5 seconds’ PH. The presented results are encouraging and show the proposed model’s clinical applicability. This study will also help practitioners in comparing the efficacy of different cueing methods.

Hashed Timelock (HTLC)-based atomic swap pro- tocols enable the exchange of coins between two or more parties without relying on a trusted entity. This protocol is like the American call option without premium. It allows the finalization of a deal within a certain period. This puts the swap initiator at liberty to delay before deciding to proceed with the deal. If she finds the deal unprofitable, she just waits for the time- period of the contract to elapse. However, the counterparty is at a loss since his assets remain locked in the contract. The best he can do is to predict the initiator’s behavior based on the asset’s price fluctuation in the future. But it is difficult to predict as cryptocurrencies are quite volatile, and their price fluctuates abruptly. We perform a game theoretic analysis of HTLC-based atomic cross-chain swap to predict whether a swap will succeed or not. From the strategic behavior of the players, we infer that this model lacks fairness. We propose Quick Swap, a two-party protocol based on hashlock and timelock that fosters faster settlement of the swap. The parties are required to lock griefing-premium along with the principal amount. If the party griefs, he ends up paying the griefing-premium. If a party finds a deal unfavorable, he has the provision to cancel the swap. We prove that Quick Swap is more participant-friendly than HTLC- based atomic swap. Our work is the first to propose a protocol to ensure fairness of atomic-swap in a cyclic multi-party setting. (Accepted in Special Session of The Fourth IEEE International Conference on Trust, Privacy and Security in Intelligent Systems, and Applications, 2022)

The numerous possible applications for Bluetooth Low Energy (BLE) Beacons constantly motivate researchers to come up with new methods to optimize the utilizations of beacons and improve the positioning accuracy while maintaining the cost to benefit balance. In this study, a novel method was proposed to optimize the localization of an indoor BLE positioning system with low beacon density in a real-world environment. The proposed method combines three major machine learning concepts (Semi-Supervised Learning, Informed Machine Learning, and Soft Computing) in its analysis pipeline and presents a different perspective to approaching Genetic Fuzzy Systems (GFSs). The presented method was competitive when benchmarked against a fully connected neural network and a kNN model.