Fuqian Shi

Indian Sign Language Recognition Using Optimized Neural Networks

Recognition of sign languages has gained reasonable interest by the researchers in the last decade. An accurate sign language recognition system can facilitate more accurate communication of deaf and dumb people. The wide variety of Indian Sign Language (ISL) led to more challenging learning process. In the current work, three novel methods was reported to solve the problem of recognition of ISL gestures effectively by combining Neural Network (NN) with Genetic Algorithm (GA), Evolutionary algorithm (EA) and Particle Swarm Optimization (PSO) separately to attain novel NN-GA, NN-EA and NN-PSO methods; respectively. The input weight vector to the NN has been optimized gradually to achieve minimum error. The proposed methods performance was compared to NN and the Multilayer Perceptron Feed-Forward Network (MLP-FFN) classifiers. Several performance metrics such as the accuracy, precision, recall, F-measure and kappa statistic were calculated. The experimental results established that the proposed algorithm achieved considerable improvement over the performance of existing works in order to recognize ISL gestures. The NN-PSO outperformed the other approaches with 99.96 accuracy, 99.98 precision, 98.29 recall, 99.63 F-Measure and 0.9956 Kappa Statistic.

Evolutionary framework for coding area selection from cancer data

AbstractCancer data analysis is significant to detect the codes that are responsible for cancer diseases. It is significant to find out the coding regions from diseases infected biological data. The infected data will be helpful to design proper drugs and will be supportable in laboratory assessments. Codes bear specific meaning on various features as well as symptoms of diseases. Coding of biological data is a key area to get exact information on animals to discover the desired medicine. In the current work, four different machine learning approaches such as support vector machine (SVM), principal component analysis (PCA) technique, neural mapping skyline filtering (NMSF) and Fisher’s discriminant analysis (FDA) were applied for data reduction and coding area selection. The experimental analysis established that the SVM outperforms PCA and FDA. However, due to the mapping facility, NMSF outperforms SVM. Thus, the NMSF achieved the preeminent results among the four techniques. Matthews’s correlation coefficient was used to evaluate the accuracy, specificity, sensitivity, F-measures and error rate of the four methods that are used to determine the coding area. Detailed experimental analysis included comparison study among the four classifiers for the deoxyribonucleic acid dataset.

Rule-based back propagation neural networks for various precision rough set presented KANSEI knowledge prediction: a case study on shoe product form features extraction

Abstract Nonlinear operators for KANSEI evaluation dataset were significantly developed such as uncertainty reason techniques including rough set, fuzzy set and neural networks. In order to extract more accurate KANSEI knowledge, rule-based presentation was concluded a promising way in KANSEI engineering research. In the present work, variable precision rough set was applied in rule-based system to reduce the complexity of the knowledge database from normal item dataset to high frequent rule set. In addition, evidence theory’s reliability indices, namely the support and confidence for rule-based knowledge presentation, were proposed by using back propagation neural network with Bayesian regularization algorithm. The proposed method was applied in shoes KANSEI evaluation system; for a certain KANSEI adjective, the key form features of products were predicted. Some similar algorithms such as Levenberg–Marquardt and scaled conjugate gradient were also discussed and compared to establish the effectiveness of the proposed approach. The experimental results established the effectiveness and feasibility of the proposed algorithms customized for shoe industry, where the proposed back propagation neural network/Bayesian regularization approach achieved superior performance compared to the other algorithms in terms of the performance, gradient, Mu, Effective number of parameter, and the sum square parameter in KANSEI support and confidence time series prediction.

Modified cuckoo search algorithm in microscopic image segmentation of hippocampus

Microscopic image analysis is one of the challenging tasks due to the presence of weak correlation and different segments of interest that may lead to ambiguity. It is also valuable in foremost meadows of technology and medicine. Identification and counting of cells play a vital role in features extraction to diagnose particular diseases precisely. Different segments should be identified accurately in order to identify and to count cells in a microscope image. Consequently, in the current work, a novel method for cell segmentation and identification has been proposed that incorporated marking cells. Thus, a novel method based on cuckoo search after pre‐processing step is employed. The method is developed and evaluated on light microscope images of rats’ hippocampus which used as a sample for the brain cells. The proposed method can be applied on the color images directly. The proposed approach incorporates the McCullochs method for lévy flight production in cuckoo search (CS) algorithm. Several objective functions, namely Otsus method, Kapur entropy and Tsallis entropy are used for segmentation. In the cuckoo search process, the Otsus between class variance, Kapurs entropy and Tsallis entropy are employed as the objective functions to be optimized. Experimental results are validated by different metrics, namely the peak signal to noise ratio (PSNR), mean square error, feature similarity index and CPU running time for all the test cases. The experimental results established that the Kapurs entropy segmentation method based on the modified CS required the least computational time compared to Otsus between‐class variance segmentation method and the Tsallis entropy segmentation method. Nevertheless, Tsallis entropy method with optimized multi‐threshold levels achieved superior performance compared to the other two segmentation methods in terms of the PSNR.

Image Fusion Incorporating Parameter Estimation Optimized Gaussian Mixture Model and Fuzzy Weighted Evaluation System: A Case Study in Time-Series Plantar Pressure Data Set

The key issue in image fusion is the process of defining evaluation indices for the output image and for multi-scale image data set. This paper attempted to develop a fusion model for plantar pressure distribution images, which is expected to contribute to feature points construction based on shoe-last surface generation and modification. First, the time series plantar pressure distribution image was preprocessed, including back removing and Laplacian of Gaussian (LoG) filter. Then, discrete wavelet transform and a multi-scale pixel conversion fusion operating using a parameter estimation optimized Gaussian mixture model (PEO-GMM) were performed. The output image was used in a fuzzy weighted evaluation system, that included the following evaluation indices: mean, standard deviation, entropy, average gradient, and spatial frequency; the difference with the reference image, including the root mean square error, signal to noise ratio (SNR), and the peak SNR; and the difference with source image including the cross entropy, joint entropy, mutual information, deviation index, correlation coefficient, and the degree of distortion. These parameters were used to evaluate the results of the comprehensive evaluation value for the synthesized image. The image reflected the fusion of plantar pressure distribution using the proposed method compared with other fusion methods, such as up-down, mean-mean, and max-min fusion. The experimental results showed that the proposed LoG filtering with PEO-GMM fusion operator outperformed other methods.

Light microscopy image de-noising using optimized LPA-ICI filter

Microscopic images are often corrupted by noise, where Poisson noise is one of the major types that can damage them. The local polynomial approximation (LPA) filter supported by the intersection confidence interval (ICI) rule was considered as an efficient filter for image de-noising. However, this filter depends on several parameters that affect its performance. In order to determine the optimal parameters, the present study employed the classic LPA-ICI (C-LPA-ICI) filter supported by optimization algorithms, namely the genetic algorithm (GA) and the particle swarm optimization (PSO) in the context of light microscopy imaging systems. Nevertheless, inclusion of the optimization algorithms increased the computational time. A novel automatic technique entitled “Standard Optimized LPA-ICI” (SO-LPA-ICI) is proposed. In this context, the average of the optimized ICI parameters was calculated, which obtained from both LPA-ICI-based GA (G-LPA-ICI) and LPA-ICI-based PSO (P-LPA-ICI). Thus, the proposed SO-LPA-ICI is included the optimal ICI parameters without optimization iterations. This procedure is proposed to speed up the optimized filter. A pool of 50 rats’ renal microscopic images is involved to test the proposed approach. A comparative study was conducted to compare the effectiveness of the four methods, namely C-LPA-ICI, G-LPA-ICI, P-LPA-ICI, and the SO-LPA-ICI for de-noising in the presence of Poisson noise. The experimental results established the outstanding performance of the SO-LPA-ICI in terms of the PSNR, MAE, and MSSIM with 28.27, 7.65, and 0.93 values, respectively. In addition, the proposed approach achieved fast de-noising compared to the G-LPA-ICI and the P-LPA-ICI.

Clinical application of modified bag-of-features coupled with hybrid neural-based classifier in dengue fever classification using gene expression data

Dengue fever detection and classification have a vital role due to the recent outbreaks of different kinds of dengue fever. Recently, the advancement in the microarray technology can be employed for such classification process. Several studies have established that the gene selection phase takes a significant role in the classifier performance. Subsequently, the current study focused on detecting two different variations, namely, dengue fever (DF) and dengue hemorrhagic fever (DHF). A modified bag-of-features method has been proposed to select the most promising genes in the classification process. Afterward, a modified cuckoo search optimization algorithm has been engaged to support the artificial neural (ANN-MCS) to classify the unknown subjects into three different classes namely, DF, DHF, and another class containing convalescent and normal cases. The proposed method has been compared with other three well-known classifiers, namely, multilayer perceptron feed-forward network (MLP-FFN), artificial neural network (ANN) trained with cuckoo search (ANN-CS), and ANN trained with PSO (ANN-PSO). Experiments have been carried out with different number of clusters for the initial bag-of-features-based feature selection phase. After obtaining the reduced dataset, the hybrid ANN-MCS model has been employed for the classification process. The results have been compared in terms of the confusion matrix-based performance measuring metrics. The experimental results indicated a highly statistically significant improvement with the proposed classifier over the traditional ANN-CS model.

Wireless Capsule Gastrointestinal Endoscopy: Direction-of-Arrival Estimation Based Localization Survey

One of the significant challenges in capsule endoscopy (CE) is to precisely determine the pathologies location. The localization process is primarily estimated using the received signal strength (RSS) from sensors in the capsule system through its movement in the gastrointestinal (GI) tract. Consequently, the wireless CE (WCE) system requires improvement to handle the lack of the capsule instantaneous localization information and to solve the relatively low transmission data rate challenges. Furthermore, the association among the capsules transmitter position, capsule location, signal reduction, and the capsule direction should be assessed. These measurements deliver significant information for the instantaneous capsule localization systems based on time-of-arrival approach, phase difference of arrival, RSS, electromagnetic, direction of arrival (DOA), and video tracking approaches are developed to locate the WCE precisely. This review introduces the acquisition concept of the GI medical images using the endoscopy with a comprehensive description of the endoscopy system components. Capsule localization and tracking are considered to be the most important features of the WCE system, thus this paper emphasizes the most common localization systems generally, highlighting the DOA-based localization systems and discusses the required significant research challenges to be addressed.

Histogram Thresholding in Image Segmentation: A Joint Level Set Method and Lattice Boltzmann Method Based Approach

The level set method (LSM) has been widely utilized in image segmentation due to its intrinsic nature which sanctions to handle intricate shapes and topological changes facilely. The current work proposed an incipient level set algorithm, which uses histogram analysis in order to efficiently segmenting images. The computational intricacy of the proposed LSM is greatly reduced by utilizing the highly parallelizable lattice Boltzmann method (LBM). The incipient algorithm is efficacious and highly parallelizable. Recently, with the development of high dimensional astronomically an immense-scale images contrivance, the desideratum of expeditious and precise segmentation methods is incrementing. The present work suggested a histogram analysis based level set approach for image segmentation. Experimental results on real images demonstrated the performance of the proposed method. It is established that the proposed segmentation methods using Level set methods for image segmentation achieved 0.92 average similarity value and average 1.35 s to run the algorithm, which outperformed Li method for segmentation.

Curvature manipulation of the spectrum of Valence-Arousal-related fMRI dataset using Gaussian-shaped Fast Fourier Transform and its application to fuzzy KANSEI adjectives modeling

ValenceArousal is regarded as a reflection of KANSEI adjectives, which is the core concept in the theory of emotional dimensions for brain recognition. This paper presents a novel method for determining the characteristics of Valence-Arousal-based timing signals using Power Spectrum Density (PSD) of fMRI images, and Gaussian filtering, segmenting, and Gaussian-shaped Fast Fourier Transform (FFT) will be applied for reprocessing fMRI images; the timing characteristics of the fMRI image signals were extracted under short-term emotional picture stimuli (within 6s). To reduce the computational complexity, a cubic curve fitting method was used to smooth the ValenceArousal timing curve, and the coefficients of the fitted curve, the mean, and the standard deviation were derived from the Gaussian-shaped Affective Norm English Words (ANEW) system, subsequently, these parameters were selected to create a 4-INPUT 2-OUTPUT TakagiSugeno (TS) type Adaptive Neuro Fuzzy Inference System (ANFIS). In the experimental study, an fMRI data-set was acquired for KANSEI-kindness picture stimuli and the FIS prediction was 0.05 less than the Root Mean Square Error (RMSE) after 24/18 iteration epochs for Valence/Arousal. These experiments showed that the proposed method effectively simplified high complexity when calculating fMRI images. The cubic curve fitting method extracted the characteristics of the ValenceArousal time series-based curves effectively and established the KANSEI adjective content more accurately by comparing with the ANEW system of ValenceArousal values. The proposed curve generation methods for the ValenceArousal response of KANSEI adjectives will be a potential application for attention-oriented product design fields.