Using cascade CNN-LSTM-FCNs to identify AI-altered video based on eye state sequence

Deep learning is notably successful in data analysis, computer vision, and human control. Nevertheless, this approach has inevitably allowed the development of DeepFake video sequences and images that could be altered so that the changes are not easily or explicitly detectable. Such alterations have been recently used to spread false news or disinformation. This study aims to identify Deepfaked videos and images and alert viewers to the possible falsity of the information. The current work presented a novel means of revealing fake face videos by cascading the convolution network with recurrent neural networks and fully connected network (FCN) models. The system detection approach utilizes the eye-blinking state in temporal video frames. Notwithstanding, it is deemed challenging to precisely depict (i) artificiality in fake videos and (ii) spatial information within the individual frame through this physiological signal. Spatial features were extracted using the VGG16 network and trained with the ImageNet dataset. The temporal features were then extracted in every 20 sequences through the LSTM network. On another note, the pre-processed eye-blinking state served as a probability to generate a novel BPD dataset. This newly-acquired dataset was fed to three models for training purposes with each entailing four, three, and six hidden layers, respectively. Every model constitutes a unique architecture and specific dropout value. Resultantly, the model optimally and accurately identified tampered videos within the dataset. The study model was assessed using the current BPD dataset based on one of the most complex datasets (FaceForensic++) with 90.8% accuracy. Such precision was successfully maintained in datasets that were not used in the training process. The training process was also accelerated by lowering the computation prerequisites.

EDR-Net: Lightweight Deep Neural Network Architecture for Detecting Referable Diabetic Retinopathy

Present architecture of convolution neural network for diabetic retinopathy (DR-Net) is based on normal convolution (NC). It incurs high computational cost as NC uses a multiplicative weight that measures a combined correlation in both cross-channel and spatial dimension of layer's inputs. This might cause the overall DR-Net architecture to be over-parameterised and computationally inefficient. This paper proposes EDR-Net - a new end-to-end, DR-Net architecture with depth-wise separable convolution module. The EDR-Net architecture was trained with DRKaggle-train dataset (35,126 images), and tested on two datasets, i.e. DRKaggle-test (53,576 images) and Messidor-2 (1,748 images). Results showed that the proposed EDR-Net achieved predictive performance comparable with current state-of-the-arts in detecting referable diabetic retinopathy (rDR) from fundus images and outperformed other light weight architectures, with at least two times less computation cost. This makes it more amenable for mobile device based computer-assisted rDR screening applications.

Non-parametric multiple inputs prediction model for magnetic field dependent complex modulus of magnetorheological elastomer

This study introduces a novel platform to predict complex modulus variables as a function of the applied magnetic field and other imperative variables using machine learning. The complex modulus prediction of magnetorheological (MR) elastomers is a challenging process, attributable to the material's highly nonlinear nature. This problem becomes apparent when considering various possible fabrication parameters. Furthermore, traditional parametric modeling methods are limited when applied to solve larger-scale cases involving large databases. Consequently, the application of non-parametric modeling such as machine learning has gained increasing attraction in recent years. Therefore, this work proposes a data-driven approach for predicting multiple input-dependent complex moduli using feedforward neural networks. Besides excitation frequency and magnetic flux density as operating conditions, the inputs consider compositions and curing conditions represented by magnetic particle weight percentage and the curing magnetic field, respectively. Extreme learning machines and artificial neural networks were used to train the models. The simulation results obtained at various curing conditions and other inputs confirm that the predicted complex modulus has high accuracy with an R² of about 0.997, as compared to the experimental results. Furthermore, the predicted complex modulus pattern and magnetorheological effect agree with the experimental data using both the learned and unlearned data.

Advancement in biosensor: "Telediagnosis" and "remote digital imaging"

Current developments in sensors and actuators are heralding a new era to facilitate things to happen effortlessly and efficiently with proper communication. On the other hand, Internet of Things (IoT) has been boomed up with er potential and occupies a wide range of disciplines. This study has choreographed to design of an algorithm and a smart data-processing scheme to implement the obtained data from the sensing system to transmit to the receivers. Technically, it is called "telediagnosis" and "remote digital monitoring," a revolution in the field of medicine and artificial intelligence. For the proof of concept, an algorithmic approach has been implemented for telediagnosis with one of the degenerative diseases, that is, Parkinson's disease. Using the data acquired from an improved interdigitated electrode, sensing surface was evaluated with the attained sensitivity of 100 fM (n = 3), and the limit of detection was calculated with the linear regression value coefficient. By the designed algorithm and data processing with the assistance of IoT, further validation was performed and attested the coordination. This proven concept can be ideally used with all sensing strategies for immediate telemedicine by end-to-end communications.

Biosensing human blood clotting factor by dual probes: Evaluation by deep long short-term memory networks in time series forecasting

Artificial intelligence of things (AIoT) has become a potential tool for use in a wide range of fields, and its use is expanding in interdisciplinary sciences. On the other hand, in a clinical scenario, human blood-clotting disease (Royal disease) detection has been considered an urgent issue that has to be solved. This study uses AIoT with deep long short-term memory networks for biosensing application and analyzes the potent clinical target, human blood clotting factor IX, by its aptamer/antibody as the probe on the microscaled fingers and gaps of the interdigitated electrode. The earlier results by the current-volt measurements have shown the changes in the surface modification. The limit of detection (LOD) was noticed as 1 pM with the antibody as the probe, whereas the aptamer behaved better with the LOD at 100 fM. The time-series predictions from the AIoT application supported the obtained results with the laboratory analyses using both probes. This application clearly supports the results obtained from the interdigitated electrode sensor as aptamer to be the better option for analyzing the blood clotting defects. The current study supports a great implementation of AIoT in sensing application and can be followed for other clinical biomarkers.

Distance Evaluated Simulated Kalman Filter with State Encoding for Combinatorial Optimization Problems

Simulated Kalman Filter (SKF) is a population-based optimization algorithm which exploits the estimation capability of Kalman filter to search for a solution in a continuous search space. The SKF algorithm only capable to solve numerical optimization problems which involve continuous search space. Some problems, such as routing and scheduling, involve binary or discrete search space. At present, there are three modifications to the original SKF algorithm in solving combinatorial optimization problems. Those modified algorithms are binary SKF (BSKF), angle modulated SKF (AMSKF), and distance evaluated SKF (DESKF). These three combinatorial SKF algorithms use binary encoding to represent the solution to a combinatorial optimization problem. This paper introduces the latest version of distance evaluated SKF which uses state encoding, instead of binary encoding, to represent the solution to a combinatorial problem. The algorithm proposed in this paper is called state-encoded distance evaluated SKF (SEDESKF) algorithm. Since the original SKF algorithm tends to converge prematurely, the distance is handled differently in this study. To control and exploration and exploitation of the SEDESKF algorithm, the distance is normalized. The performance of the SEDESKF algorithm is compared against the existing combinatorial SKF algorithm based on a set of Traveling Salesman Problem (TSP).      

Feature selection using angle modulated simulated Kalman filter for peak classification of EEG signals

In the existing electroencephalogram (EEG) signals peak classification research, the existing models, such as Dumpala, Acir, Liu, and Dingle peak models, employ different set of features. However, all these models may not be able to offer good performance for various applications and it is found to be problem dependent. Therefore, the objective of this study is to combine all the associated features from the existing models before selecting the best combination of features. A new optimization algorithm, namely as angle modulated simulated Kalman filter (AMSKF) will be employed as feature selector. Also, the neural network random weight method is utilized in the proposed AMSKF technique as a classifier. In the conducted experiment, 11,781 samples of peak candidate are employed in this study for the validation purpose. The samples are collected from three different peak event-related EEG signals of 30 healthy subjects; (1) single eye blink, (2) double eye blink, and (3) eye movement signals. The experimental results have shown that the proposed AMSKF feature selector is able to find the best combination of features and performs at par with the existing related studies of epileptic EEG events classification.

Evaluation of different time domain peak models using extreme learning machine-based peak detection for EEG signal

Various peak models have been introduced to detect and analyze peaks in the time domain analysis of electroencephalogram (EEG) signals. In general, peak model in the time domain analysis consists of a set of signal parameters, such as amplitude, width, and slope. Models including those proposed by Dumpala, Acir, Liu, and Dingle are routinely used to detect peaks in EEG signals acquired in clinical studies of epilepsy or eye blink. The optimal peak model is the most reliable peak detection performance in a particular application. A fair measure of performance of different models requires a common and unbiased platform. In this study, we evaluate the performance of the four different peak models using the extreme learning machine (ELM)-based peak detection algorithm. We found that the Dingle model gave the best performance, with 72 % accuracy in the analysis of real EEG data. Statistical analysis conferred that the Dingle model afforded significantly better mean testing accuracy than did the Acir and Liu models, which were in the range 37–52 %. Meanwhile, the Dingle model has no significant difference compared to Dumpala model.