Recent Patents on Computer Science - Volume 11, Issue 1, 2018

Background: The massive database of videos is growing day by day in this era. Analyzing such huge data is always a time-consuming process. The effective use of video content requires a user-friendly access to information. This leads to the evolution of the research area known as video summarization. The effective techniques of video summarization, the videos have let to analyze the content of large volumes of digital video sequences in various categories, such as surveillance, documentaries, movies, sports, lectures, and news. In video summarization, the automatic selection of necessary and informative section from videos using accurate algorithms is essential. The keyframe extraction in video summarization is intended to suffice comprehensive analysis of video by eliminating replications and extraction of keyframes from the video. Methods: Recent keyframe extraction techniques like clustering, shot, visual content based keyframe extraction methods are discussed for effective keyframe extraction. Results: First an introduction of various techniques for keyframe extraction pursued by the state-ofthe- art review on their properties. Although we have outlined some ideas for effective evaluation of video keyframes, the analytical evaluation of various keyframe extraction techniques is discussed and the approaches based on the methods, dataset and the results are compared. Conclusion: In the recent years, the use of digital video data has been increasing significantly due to the extensive use of multimedia applications in the areas of education, entertainment, business. So the video has received an incredible attention and research interest in video processing. The use of keyframe extraction has been given incredible attention, in this work, we have carried out a comprehensive survey and review of the research in keyframe extraction techniques. We believe the review paper will provide an update for the reader regarding the progress of keyframe extraction by different keyframe extraction techniques.

Background: The rapid improvement in technology enables an Electroencephalogram (EEG) to detect a diverse range of brain disorders easily. The design of sophisticated signal processing methods for an efficient analysis of the EEG signals is exceptionally essential. Raw EEG signal is contaminated by noise and artefacts that modify the spectral-spatial and temporal information of the signal and renders inaccurate clinical interpretation. Denoising of the signal is the first step to refine the signal quality and identify patient's mental state from the signal although it is not an easy task because of high dimensionality and complexity of EEG signal. The present study highlights three conditions of the brain namely stroke, brain death, and a healthy state. The primary concern is to detect the most abnormal conditions of the brain, i.e., an EEG with a critical stage. Method: This paper introduces a neoteric technique for the analysis of EEG signals of the three conditions using filters such as Fuzzy filter and wavelet orthogonal filter to obtain highly accurate resultant signals. Further, the resultant filter is trained in Neural Network for predicting the brain abnormalities. The proposed system is found to be efficient in denoising the EEG waves. Results: The result shows that the classification accuracy of multiclass EEG dataset achieved and the performance of ANN is high and it was found to be the best validation performance of ANN which is 0.2303. Conclusion: This paper comprehensively describes the denoising of the EEG signals that will provide accuracy in the diagnosis of the EEG to detect brain disorders. The Fuzzy filter pre-processes the signals by considering the noisy signal by an ideal value in such a way that the desired metric (the filtered output) is reduced. The orthogonal wavelet filter produces a single scaling function and wavelet function. The EEG features are extracted from multiple-level decompositions of EEGs by DWT. Finally, the features are classified using Back propagation artificial neural network that categorizes the EEGs to make the diagnosis easier for the brain abnormalities.

Background: Generally, it is difficult to obtain the literature metadata in a unified way because the source data of literature is heterogeneous. Researchers developed a series of systems marked by digital object to manage them and obtained a good effect. Though there are several DOI systems, we face with some problems in promoting the use of them. Objective: To address this issue of promoting literature identifier extraction, this paper has proposed automatic literature metadata extraction from DataCite services. Method: This paper describes Patent Publication Number CN103279361A, titled "Method and System for Bookmark-triggered Literature Sharing", issued by State Intellectual Property Office of the P.R.C. on January 27, 2016. A literature metadata extraction system supporting both personal computer and mobile terminal is developed using the integration of DataCite content negotiation, DataCite metadata search, and HTML template extraction. The architecture of this system is divided into model, view, service and controller. An important contribution of this article is to design a cross-platform and universal way to extract digital literature with/without DOI. Results: The analysis of application's effect and piratical test case show the ability to verify the authenticity of automatic literature metadata extraction from DataCite services. The contributions of our method are literature identifier extraction from DOI proxy, template extraction using Roadrunner, and bookmarklet-based literature sharing. Conclusion: The idea and a disclosed embodiment of a patent (Patent CN103279361A, issued by State Intellectual Property Office of the P.R.C.) are presented, which is based on the distribution of literature metadata extraction. In one disclosed embodiment, this method contains literature identifier extraction from DOI proxy, template extraction using Roadrunner, and bookmarklet-based literature sharing.

Background: The most important question about the application and development of cloud computing is the privacy risk, as it has become the key factor for users in selecting cloud services. In order to have a better understanding of the privacy risk, resolve or make it controllable in an acceptable range, risk assessment is very necessary. However, specialized research on the assessment has not yet been found. Methods: This paper has used Information Entropy Theory and Fuzzy Set Theory to evaluate cloud computing privacy security risk. Results: This paper analyzes the research actuality of cloud computing privacy security and combs a number of cloud computing privacy security risks on the basis of digging deep into the characteristics of cloud computing and previous studies, thus further establishing an evaluation index system for privacy risk initiatively. Then, the entropy weight theory and fuzzy set theory are used to establish a fusion of multiple factors of the assessment model. Conclusion: A practical case is given to prove the feasibility, rationality and effectiveness of the evaluation method.

Background: Deep Neural Network (DNN) has attracted great attention in regression and classification problems. However, the traditional DNN fails to provide favorable prediction performance owing to their inherent shortcomings of the Back Propagation (BP) algorithm, such as slow convergence and local optimum. To solve this problem, a novel DNN algorithm called Multi-Layer Echo State Network (ML-ESN) is proposed. Method: This method utilizes Echo State Network (ESN) based unsupervised and supervised learning process. After discussing some related patents and methods, the ML-ESN structure is given. In the ML-ESN with M+1 hidden layers, ESN based Auto-Encoder (ESN-AE) is employed by the front M hidden layers for feature extraction in the unsupervised learning process, which can effectively extract abstract features from the original data. To overcome the under-fitting and over-fitting problem of ESN based classifier, the Penalty Factor ESN (PF-ESN) is presented to act as classifier by the M+1th hidden layer in the supervised learning process. Results: The experiments demonstrate that ML-ESN outperforms ESN, Tikhonov regularization ESN (TR-ESN) and some DNN models such as Deep Belief Network (DBN) and stacked auto encoder (SAE). Owing to the fast learning speed of ESN, the training time consumtion of ML-ESN is also shorter than DBN and SAE. Conclusion: The method provides references for fast and high accuracy classification and feature extraction.

Background: Neural networks have been very successful in various sorts of aspects, such as computer vision, intelligent prediction and control, image processing, and natural language processing. Coupled delayed neural networks of neutral-type possess more sophisticated actions than a single node neural network. Nevertheless, it is still a challenge to consider the synchronization issue of coupled neural networks of neutral-type processed unknown transition rates. Method: The authors in this paper propose an augmented Lyapunov-Krasovskii functional and use further improved integral inequality and Dynkin's formula to acquire the delay-dependent cluster synchronization criteria that comprise upper and lower bounds of delays, which have less conservative criteria than the existing results. Meanwhile, the transition rates in the Markov jump coupled delayed neural networks of neutral-type which are not fully known and the coupling configuration matrices are not confined to symmetry. Hence, the matrices in this system have less constrained conditions than most of the existed papers. Result: By applying the infinitesimal operator and the special matrix K , the newly delaydependent cluster synchronization criteria are presented for the Markov jump coupled neural networks of neutral-type with unknown transition rates. Conclusion: The less conservative cluster synchronization criteria have been proposed for Markov jump coupled neutral-type networks with unknown transition rates by adopting the augmented Lyapunov- Krasovskii functional and integral inequality. A numerical example has been provided to illustrate the effectiveness of the proposed approach. The method in this paper is less conservative than the existed ones.

Objective: To investigate the impact of sarcasm in analyzing the sentiments from tweets. Design: 1. The Tweets related to five different domains are collected from the Twitter by creating Twitter developer account. 2. The Tweets are preprocessed in order to extract the features (Term Frequency, Entropy, Gain Ratio) from the Tweets. 3. Proposed an Iterative algorithm in updating the dictionary with Negative Phrases and sentiment words. 4. Assigned a polarity to each tweet using the Dictionary based approach. 5. Tweets with Zero scores are detected as Sarcasm tweets. 6. Analysis on Variance (AOV) Test is performed on the scores obtained. 7. Perform prediction on the scores using Machine Learning Algorithms. 8. Estimated the Mean Square Prediction Error (MSPE) using Cross Validation. Outcome: The impact of sarcasm on sentiment analysis is measured in terms of Precision, Recall and F-score. Results: 1. Scores of Tweets based on sentiment words and Negative Phrases. 2. Summary of Analysis on variance on Scores obtained. 3. Performance of Machine Learning Algorithms in Detecting the sarcasm from tweets. Conclusion: This paper has been presented to put forth the hypothesis that changes sentiment polarity (positive to negative) of sentences can be used as a feature for detecting sarcasm within product review length bodies of text. With the level of accuracy achieved, this scoring technique can be implemented as software for social networks. Also, these efforts may be a useful tool for learning about patterns in sarcasm or making better dictionaries of offensive words.

Background: The fault of a hydroelectric generating unit is mostly expressed in the form of vibration, and the reason is very complicated. Therefore, it is difficult to describe the mapping relationship between the fault cause and fault symptom using the traditional approach. Methods: To improve the accuracy of fault diagnosis for a hydroelectric generating unit, we proposed a hybrid intelligent diagnosis technology in which the BP neural network is trained by cuckoo search algorithm with a quantum mechanism (QCSBP). Results: Through the experimental study, we demonstrate that cuckoo search with a quantum mechanism (QCS) is superior to the five comparable approaches, and the proposed QCSBP model has the highest diagnostic accuracy. Conclusion: The QCSBP model can effectively identify the fault state of a hydroelectric generating unit, and is a fault diagnosis method with application prospect.