Current Issue : January-March Volume : 2024 Issue Number : 1 Articles : 5 Articles
Errors in analog-to-digital conversion (ADC) occur due to internal links or other electronic parts; faults that may occur during code conversion cannot be overlooked because signal digitalisation demands a large dynamic range and high resolution. This paper presents a new and accurate self-test method to compensate for one of the most effective errors of ADC because of its effect, which may result in a missing code, which is a differential nonlinear (DNL) of a 10-bit SAR-ADC. The proposed method includes three stages: DNL error modelling for nonideal system implementation, detection, and correction. To evaluate the proposed technique, sinusoidal and sawtooth signals are applied as analog inputs to the proposed system. Adaptivity, speed, and accuracy are the main motivations of this work, which provide high accuracy compared to other techniques, up to 9.6 ENOB and 59.2 SNR with sawtooth signal and 9.5ENOB and 59.2 SNR with sinewave signals....
The use of short message service (SMS) and e-mail have increased too much over the last decades. 80% of people do not read emails while 98% of cell phone users daily read their SMS. However, these communication media are unsafe and can produce malicious attacks called spam. The e-mails that pretend to be from a trusted company to provide “financial or personal information” are phishing e-mails. These e-mails contain some links; users might download malicious software on their computers when they click on them. Most techniques and models are developed to automatically detect these “SMS and e-mails” but none of them achieved 100% accuracy. In previous studies using machine learning (ML), spam detection using a small dataset has resulted in lower accuracy. To counter this problem, in this paper, multiple classifiers of ML and a classifier of deep learning (DL) were applied to the SMS and e-mail dataset for spam detection with higher accuracy. After conducting experiments on the real dataset, the researchers concluded that the proposed system performed better and more accurately than previously existing models. Specifically, the support vector machine (SVM) classifier outperformed all others. These results suggest that SVM is the optimal choice for classification purposes....
Implementing an efficient system for emotion recognition has recently posed a challenge that has not been fully developed yet. Facial emotion recognition (FER) is an important subject matter in the fields of artificial intelligence (AI) since it exhibits a greater commercial potential. This technique is used to analyse various sentiments and reveal a person’s behavior. It could be related to the mental or physiological state of mind. This paper mainly focuses on a human emotion recognition system through a detected human face. Its accuracy was improved via different data augmentation tools, early stopping, and generative adversarial networks (GANs). Compared to previous methods, experimental results show that the proposed method provides a 0.55% to 35.7% gain performance....
Earthquake-induced soil liquefaction (EISL) can cause significant damage to structures, facilities, and vital urban arteries. Thus, the accurate prediction of EISL is a challenge for geotechnical engineers in mitigating irreparable loss to buildings and human lives. This research aims to propose a binary classification model based on the hybrid method of a wavelet neural network (WNN) and particle swarm optimization (PSO) to predict EISL based on cone penetration test (CPT) results. To this end, a well-known dataset consisting of 109 datapoints has been used. The developed WNN-PSO model can predict liquefaction with an overall accuracy of 99.09% based on seven input variables, including total vertical stress (σv), effective vertical stress (σ v ), mean grain size (D50), normalized peak horizontal acceleration at ground surface (αmax), cone resistance (qc), cyclic stress ratio (CSR), and earthquake magnitude (Mw). The results show that the proposed WNN-PSO model has superior performance against other computational intelligence models. The results of sensitivity analysis using the neighborhood component analysis (NCA) method reveal that among the seven input variables, qc has the highest degree of importance and Mw has the lowest degree of importance in predicting EISL....
Sleep apnea is one of the most common sleep disorders in the world. It is a common problem for patients to suffer from sleep disturbances. In this paper, we propose a deep convolutional neural network (CNN) model based on the oxygen saturation (SpO2) signal from a smart sensor. This is the reason why we called ZleepNet a network for sleep apnea detection. The proposed model includes three convolutional layers, which include ReLu activation function, 2 dense layers, and one dropout layer for predicting sleep apnea. In this proposed model, the use of signals for detecting the sleep apnea can be reduced from 25 sensors to 1 sensor. We conducted experiments to evaluate the performance of the proposed CNN using real patient data and compared them with traditional machine learning methods such as least discriminant analysis (LDA) and support vector machine (SVM), baggy representation tree, and artificial neural network (ANN) on publicly available sleep datasets using the same parameter setting. The results show that the proposed model outperformed the other methods with the accuracy of 91.30% with the split rate of 0.2% in which the training data are 20% and testing data are 80%. The accuracy of the proposed CNN is 90.33% when compared with the LDA which achieved 86.5% accuracy with the split rate of 0.5% in which training data are 50% and testing data are 50%. It achieved 91.56% accuracy when compared with the support vector machine (SVM) in which training data are 70% and testing data are 30%. The achieved accuracy of the proposed CNN is 91.89% when compared with bagging representation tree in which training data are 90% and testing data are 10%. The accuracy of the proposed CNN is 91.30% in which training data are 83% and testing data are 17% when compared with artificial neural networks (ANN)....
Loading....