Anasua Sarkar

Class Discriminator-Based EMG Classification Approach for Detection of Neuromuscular Diseases Using Discriminator-Dependent Decision Rule (D3R) Approach

Classification of EMG signals is essential for diagnosis of motor neuron diseases like neuropathy and myopathy. Although a number of strategies have been implemented for classification, none of them are efficient enough to be implemented in clinical environment. In the present study, we use ensemble approach of support vector machines for classification of three classes (normal, myopathic and neuropathic) of clinical electromyogram (EMG). Our proposed approach uses time and time–frequency features extracted from EMG signals. By employing two types of feature set for same class discriminators, we are able to select the best feature set-discriminator pairs. The decision made by each selected classifier is used to generate the final class for an input EMG signal through majority voting. Our proposed method yields higher accuracy of 94.67% over 89.67% for multiclass SVM classifier.

Time domain multi-feature extraction and classification of human hand movements using surface EMG

Hand movement recognition from electromyogram (EMG) signals is a crucial element for design of electrically controlled limb prostheses and for developing human computer interfaces (HCIs). The study focuses on extraction of significant features from raw EMG signals corresponding to six different grasping hand movements and then using them to classify the respective hand movements using different classifiers. We propose a hybrid multi-feature set comprising of Autoregressive (AR), Root Mean Square (RMS), Zero Crossing (ZC), Slope Sign Change (SSC), Waveform Length (WL) and Mean Absolute Value (MAV) time domain features. We use four different classifiers (k-NN, LDA, QDA and Subspace Discriminant Ensemble) for this experiment. Different set of features yield varying accuracies depending upon the choice of classifiers. Conventional AR feature set provides maximum accuracy of 80.83% with LDA classifier. Similarly, the feature set excluding AR obtains maximum accuracy of 72.5% with k-NN classifier. Our proposed multi-feature set of all six features provides the highest accuracy of 83.33% with Ensemble classifier which is significantly higher than the accuracy values of other feature sets. We validate the results on a large dataset of 600 sEMG signals corresponding to 6 different grasping hand movements.

Improving Modified Differential Evolution for Fuzzy Clustering

Differential evolution is a real value encoded evolutionary algorithm for global optimization. It has gained popularity due to its simplicity and efficiency. Use of special kind of mutation and crossover operators differentiates it from other evolutionary algorithms. In recent times, it has been widely used in different fields of science and engineering. Among recently developed various variants of differential evolution, a modified technique called Modified Differential Evolution based Fuzzy Clustering (MoDEFC-V1), was proposed by the authors of this article to improve the speed and accuracy of convergence of differential evolution with a new mutation operation. However, it has a certain limitation of finding global optimum value while searching in solution space. To overcome the limitation of MoDEFC-V1, in this article, we have proposed two different improved versions of MoDEFC called MoDEFC-V2 and MoDEFC-V3 in order to do the underlying optimization such as clustering of patterns better. The effectiveness of the proposed versions is demonstrated for two synthetic and four real-life datasets. Moreover, the superiority of MoDEFC-V2 and MoDEFC-V3 is shown by comparing with state-of-the-art methods qualitatively and quantitatively. Finally, two sample independent one-tailed t-test is performed in order to judge the superiority of the results produced by the proposed versions.

QSAR Model for Mast Cell Stabilizing Activity of Indolecarboxamidotetrazole Compounds on Human Basophils

Indolecarboxamidotetrazole compounds are well known as potential anti allergic agents due to their mast cell stabilizing activity on human basophils. A quantitative structure activity relationship (QSAR) model has been generated using Multiple Linear regression (MLR) for the prediction of inhibition efficiency of indolecarboxamidotetrazole derivatives. Twenty-one compounds with their activities expressed as % inhibition (PI) are collected. Descriptors are generated using Chemistry Development Kit. Three models are built and the models are evaluated using multiple correlation coefficient (R) and residual standard deviation (s). Considering the quality and accuracy of the predicted models, model 1 is the best, because it predicts biological activity which is almost closed to that of experimental value. This model is externally validated. This built model can be used to calculate inhibition efficiency of natural mast cell stabilizers containing caroxamidotetrazoles as antiallergic chemical in future.

Automatic mixed pixel detection using a new hybrid Cellular automata approach on satellite image

Mixed-pixels classification in land-cover regions is a challenging task in remote sensing imagery. To classify mixed-pixels, vagueness is always the main characteristic by handling uncertainty. We propose a hybrid approach for pixel classification using Rough sets and Cellular automata models to solve this problem. Multiple belongingness and vagueness among data can be handled efficiently using Rough set theory and is appropriate for detecting arbitrarily-shaped clusters in satellite images. We propose a rough-set based automatic heuristically decision-rule generation algorithm to obtain initial set of clusters. As a discrete, dynamical system, cellular automaton comprises of uniformly interconnected cells with states. In the second phase of our method, we utilize a 2-dimensional cellular automaton to prioritize allocations of mixed pixels among overlapping land cover regions. We experiment our algorithm on Ajoy river catchment area. The segmented regions are compared with well-known FCM and K-Means methods and the ground truth knowledge, which shows superiority of our new approach.

Quality evaluation of wavelet functions for myopulse suppression in electrocardiogram

ECG is susceptible to parasitic myopulses due to the overlapping frequency bandwidth of ECG and EMG. EMG signal has a bandwidth of about 20–500 Hz and overlaps with the ECG frequency range. i.e. 0.05–150 Hz. These interferences occur due to movement of muscles and respiratory actions during ECG recording. Removal of EMG noise from ECG is an important criterion for proper analysis of the signal. In this study, we evaluated the denoising performance of wavelet functions by considering SNR as the quality judgement parameter. DWT provides better denoising over traditional filtering techniques. The level of decomposition plays an important role in denoising quality. There is variation in the performance of hard and soft thresholding with varying levels of decomposition. Hybrid thresholding is the best noise estimation and cancellation technique. Wavelet functions with more oscillations produce good denoising than others.