Embedded Files
(Volume: 3, Issue: 6)
Power-Efficient And Spectrally-Efficient OFDM For Fiber Optic IM/DD Systems
Fiber optic communications, wireless optical communications and Visible Light Communication (VLC) have Intensity Modulation/ Direct Detection (IM/DD) system as their core component, owing to its simple, cost-effective and high-speed optical data transmission. However, this system also has bandwidth limitations and emphasizes data communication with no noise or interference. Hence, Orthogonal Frequency Division Multiplexing (OFDM), the modulation technique known for its spectral efficiency, is greatly used in IM/DD system. With OFDM, the available bandwidth is divided into closely-spaced orthogonal subcarriers to hold individually-modulated optical data, which are parallelly-transmitted without interference. Yet, the non-linearity of the optical devices causes signal clipping distortions. Hence, DC-bias optical OFDM (DCO-OFDM) came into effect to avoid signal clipping. DCO-OFDM offered better spectral efficiency, yet it was power-inefficient as it consumed additional power for the DC- bias, which is a great concern for the battery-powered or energy-efficient applications!!! Hence, Symmetrically-Clipped Optical OFDM (SCO-OFDM) and Asymmetrically-Clipped Optical OFDM (ACO-OFDM) schemes were used to improve the power efficiency. Whenever the signal crossed the threshold, the SCO-OFDM clipped the optical signal symmetrically and the ACO-OFDM clipped the negative values of the optical signal alone to zero. Since the IM/DD systems cannot detect negative values, ACO-OFDM seemed superior for use with high power efficiency. However, spectral efficiency is an issue again because of the information loss in the even subcarriers. Many researchers have tried to improve both the power as well as the spectral efficiency in the IM/DD system, but the trade-off between the two was hard to achieve with added complexity and costs. One research in 2015 suggests both the asymmetrical and the symmetrical clipping of optical signal in OFDM, termed as ASCO-OFDM. Here, the even subcarriers and the odd subcarriers were modulated by the two-frame SCO-OFDM and the two-frame ACO-OFDM, respectively, eliminating DC bias. Though the Symbol Error Rates and power efficiency were better, the computational complexity at the receiver also increased with the larger achieved bandwidths. Hence, Mohammed Salman Baig, Mohammed Thamer Alresheedi, Mohd Adzir Mahdi and Ahmad Fauzi Abas have modified the ASCO-OFDM to incorporate spectral efficiency as well. In their article in Optical and Quantum Electronics, Springer, vol.55, the authors have used all the subcarriers as in ASCO-OFDM, except that the two-frame SCO-OFDM is replaced with a single-frame SCO-OFDM, clipping one side of symmetry and containing the clipping distortion on the other side of symmetry. This was done to estimate the clipping distortion in time domain and to ease up the data recovery at the receiver. The authors have found their approach to be 1.333 times spectrally-efficient than ASCO-OFDM with lower Peak Average Power Ratio (PAPR). However, the authors state that their research assumed only Additive White Gaussian Noise (AWGN) channel and need further investigations in multi-path channel settings. Since high-speed optical communications with the IM/ DD systems always incur multi-path channel noises and interferences, there is a great way ahead for the upcoming researchers to explore in this domain.
Machine Learning Models For Breast Cancer Survival Analysis
Breast cancers, whether in-situ or metastasized into invasive cancers, have turned out as a serious illness, predominantly affecting women of almost all age groups than men. The National Cancer Institute estimates about 3,10,720 females in U.S. to be diagnosed with breast cancer in 2024, out of which 42,250 have been estimated to die!!! So, breast cancer prognosis becomes highly-vital to understand the progression of the disease as well as its impact on the patients’ survival rates and to provide proper medication to the patient for rapid cure. Generally, the prognosis involves the assessment of various biomarkers that influence the growth of abnormal cells in the breast. Few biomarkers include, the estrogen or progesterone hormone receptors, the HER2 (Human Epidermal Growth Factor Receptor) protein or the mutation of BRCA1 and BRCA2 genes. Additionally, there are also alternative means of prognosis like, the estimation of the tumor’s size or its histologic grade and the estimation of the lymph node count being affected through metastasis. However, all these estimated prognostic features can never be the same in all patients, as they differ by age, tumor characteristics or treatment conditions. Hence, finding the most important prognostic features and their correlation can certainly influence the medical therapy and the patient survival rates, but at the cost of excessive and uncertain manual effort. So, statistical models (especially, the Cox Proportional Hazards model) have been widely used to find the prognosis feature correlations, without or less human intervention. Since these models also fail to predict high non-linear patterns at times, researchers have shown interest to incorporate machine learning and deep learning approaches into the survival analysis. Keren Evangeline I, S. P. Angeline Kirubha1 and J. Glory Precious have compared the DeepHit model against the Cox PH model and the random survival forests model in their article in Multimedia Tools and Applications, Springer, vol. 82. Using the clinical as well as the pathological features from the Molecular Taxonomy of Breast Cancer International Consortium dataset (METABRIC), the researchers have identified the age at diagnosis, estrogen as well as progesterone hormone receptors and the stage of tumor as the most significant prognostic features for predicting the survival rate of the breast cancer patient. However, the researchers also state that these significant prognostic features were identified only based on clinical or pathological covariates and not on genetic covariates. Hence, the survival analysis of breast cancer in future might consider the genetic covariates too with more-sophisticated and novel machine learning approaches. Image courtesy: www.freepik.com
ADNs Mandate Optimally-Sized DES At Optimal Locations
The desire for the sustainable and reliable distribution of power have caused the Active Distribution Network (ADN) to gain more attention nowadays. This energy system, chiefly encompassing entities like, Distributed Energy Sources (DES) and the battery storage or the power management systems, is aimed to render real-time monitoring and management of power flow within the grid using smart and automated controls. In fact, the increased penetration of DES in ADNs is an added advantage, as they tend to render cost-effective and flexible renewable power generation at multiple locations in the grid. However, not all the DES like, solar, biomass, wind, fuel cell, thermal and similar other DES of same size can operate at the same location and supply incessant power based on customer needs. Their environmental and operational settings might be cited as the major reasons behind, provoking research on optimal sizing and location-planning of DES in ADNs. Numerous researchers have attempted to optimally place and size the DES in ADNs to provide reduced power losses, balanced nodal voltage profiles and unintermittent power supply based on customer demands. Two researchers, Deepak Sharma and Pushpendra singh, too have analyzed various constraints for optimally planning the location and the size of DES, envisaging increased DES penetration at the metering ends of ADN. The constraints being analyzed by the researchers included the energy demand, carbon emission losses, active power, cost, charging status, voltage ratios in the main grid and distributed generators. Presenting their analysis on the constraints and the approaches that were so far used for active power distribution in the 2020 Second International Conference on Inventive Research in Computing Applications (ICIRCA), the researchers have elucidated the future energy engineers and planners about the economic and eco-friendly ways of integrating DES to the grid-tied microgrids.
Image courtesy: www.freepik.com
Emotion Recognition Using ECG
Recognizing emotions from facial expressions, speech or text is one of the demanding techs of this era, which seeks automation in almost all applications. However, all the mentioned features for achieving emotion recognition are subjective to interferences from their surroundings. For instance, a person might be worrying inside, but might show happiness in his/ her face or speech externally. So, is there any alternative and true feature for emotion recognition to aid in real-time applications? Of course, yes. It is the physiological signal. Generally, the physiological signals acquired from the heart or the skin have direct correlation with the emotions that they serve as rich features for emotion recognition. V. K. Patil, V. R. Pawar, S. P. Kulkarni, T. A. Mehta and N. R. Khare too have used the Electrocardiogram (ECG) signal for emotion recognition, in spite of using it for the conventional heart abnormality classification. Employing an ECG sensor, a temperature sensor and a signal processing circuitry, the researchers’ intention was to compare and find the best machine learning approach for ECG-based recognition of emotions like, happy, sad, stressed and neutral. In their article in International Journal of Engineering, vol. 36(6), the researchers have stated that their emotion-recognizing hardware can be used for parental control of games. Directing the future researchers to also use other bodily parameters such as, the skin conductance or saturation of peripheral oxygen (SpO2) for recognizing emotions, the researchers have seeded a strong notion to achieve human-computer interaction, mental health monitoring, personalized learning, gaming, automated driving, client-based recommendations, industrial automation and many more applications of this intelligent world in an improved and trustable way.
Image courtesy: www.freepik.com
JLFETs For Better Switching At Elevated Temperatures
The VLSI technology has caused the recent electronic device or circuit fabrication to integrate small or nano-sized transistors, which are heat-resistant and support high-speed operations with least power consumption as possible. The Gate-All-Around Junctionless Field Effect Transistor (GAA-JLFET) is one such transistor kind that imparts enhanced switching characteristics and power efficiency, even though it is scaled down to nano-sizes. Generally, the devices like Junction Field Effect Transistors (JFETs) produce short channel effects and leakage currents, which dwindle the switching characteristics and power efficiency, respectively. GAA-JLFETs tackle this issue with its great structure, which involve a gate material surrounding the uniformly-doped channel that is devoid of p-n junctions. Though this gating structure allows for better electrostatic control over the channel and eases fabrication, the performance enhancement over the other transistor topologies still relies on the material choice and the amount of dopant material. Additionally, the quantum tunneling in JLFETs also needs to be analyzed in detail to make it operate effectively at high breakdown voltages or temperatures. Zahied Azam and Ashok Kumar have investigated the performance of P+ SiC core-shell JLFET, as whether it can withstand high temperature based on the gate bias, doping effect and the properties of the semiconductor material. The researchers have presented their investigation results of this JFLET, involving a SiC nanowire with P+ core and N shell around, in Micro and Nanostructures, Elsevier, vol. 191. As per the researchers, the P+ SiC core-shell JLFET improved the electrostatic integrity with decreased lateral band-to-band tunneling at elevated temperatures, supporting high breakdown voltage applications. Additionally, the researchers have suggested the optimal parameters for drain current estimation too. Since nano and even more small-scaled transistors are going to be the ultimate choice of the modern space-constrained or power-constrained devices, like the IoT (Internet of Things) devices, analyzing deeper on various transistor structures or their characteristics has a great scope in future. Image courtesy: www.freepik.com
Page updated
Google Sites
Report abuse