ESTIRJ Volume 3, Issue 2 June 2019
Paper 1: BRAIN TUMOR DETECTION FROM MRI IMAGES BY FUSION OF EDGION GROWING AND EDGE DETECTION ALGORITHYM (1-6)
Abstract: An abnormal increase in tissues is known as tumor. Brain Tumor is a mass of tissue abnormal in nature which increases uncontrollably in size and multiply irrepressibly not checked by the mechanism that functions for controlling the normal cells. In this proposed system, a brain tumor is to be detected from multiple MRI images. The process involves processing an image through edge detection algorithm i.e Canny Edge Detection. The same image is segmented using Automatic Region Growing Segmentation technique. The two segmented images from edge detection and region growing techniques are processed through fusion of two segmentation results. An image fusion is a process that combines the results from two images of same brain tumor MRI into a single image. The resultant image has more information than the segmentation by respective algorithms alone. The results obtained from the fusion method show the efficiency and performance of proposed fusion approach.
Keywords: Region Growing Segmentation, Edge Detection Segmentation, Morphological Operation, Wavelet Transform Based Image Fusion
Paper 2:SPEECH EMOTION RECOGNITION USING SUPPORT VECTOR MACHINE
Abstract:The Human computer interaction (HCI) whereas emotion recognition of speech through machine learning has many outcomes like to monitor the psychological behavior of human; it is also used in lie detection. The emotion of human can be recognized by different ways mostly by speech or facial expression. To make more natural communication edge between human and computer; speech has one of the primary objectives to recognize an emotion as facial expression. But to classify the accuracy of the speech emotions, it has many flaws and should be overcome. Here in this research; to recognize the emotion of speech, we are going to use Support Vector Machine (SVM) it perfectly classifies the data into groups; that categorization is accomplished by linear or no-linear untying plane in an input features space of the data sets. Whereas, in this research four features of speech corpus are selected and extracted out those fundamental speech recognition features form each speech corpus and classified that data by using SVM and achieved the testing accuracy of 87.5% correctly. (7-12)
Keywords: Human Computer Interaction, Support Vector Machine, Speech Emotion, Pitch, Energy, Mal Frequency Cepstrum Co-Efficient, Zero-crossing
Paper 3: ANALYSIS AND REDUCTION OF 11KV DISTRIBUTION LINE LOSSES OF HESCO (13-20)
Abstract: This research is a study on reduction of energy loss in the distribution system. This research is done in the power distribution system using the PSS SINCAL / adept program as a modelling tool. A considered method for reducing technical losses adds new equipment (transformers) and increases the transmission capacity of the power distribution system by reconfiguring the network based on the optimal power flow. Causes of the loss reduction. The advantage is to reduce the energy cost of the distribution system. Profit is calculated at a cost that will indicate the economic basis for investing in a source of finance that will reduce losses. In model studies using Qasimabad Grid Station of HESCO a method of reducing energy loss based on optimal power flow, combined with cost-benefit analysis and increasing the power of the power distribution system by adding new equipment, which will explain. This method was implemented on the example of a real distribution system of one of the regions Qasimabad Grid Station HESCO. Studies have shown that a transformer can help reduce energy losses in a network. Transformer costs can be offset by reduced losses. Therefore, an additional transformer can significantly improve the performance of the distribution system and reduce energy loss. Network reconfiguration should be applied correctly, but if used improperly, there is a large loss of energy. Finally, when developing an appropriate plan, the installation of the transformer will be executed correctly and operated.
Keywords: Hyderabad Electric Supply Company; Energy Loss Reduction; High Tension; Kilo Volt; Low Tension; Pole Mounted Transformer; Geographic information system; Siemens Network Calculation
Paper 4: PERFORMANCE OF MSE WALL REINFORCED WITH CRUMB TIRE RUBBER (21-26)
Abstract: Nowadays, Walls and embankments of highway are mostly made with the help of Mechanically Stabilized Earth (MSE). In this research, crumb tire rubber (CTR) was used as a reinforcement to analyze the changes in the stability of MSE wall. The rubber tire was acquired from Tire Rubber Factory, korangi Industrial Area, Karachi while the sand used in this research was obtained from the Indus River near Jamshoro. Direct Shear Box Test, Compaction Test and Sieve Analysis was done on the STR mixed sand to obtain parameters used to model MSE wall in finite element program, PLAXIS 2D. Direct Shear Box Test was simulated in the PLAXIS 2D to check the results and verification was also done on the Chungsik’s two-tier wall. It was concluded that the 15% by weight of crumb rubber tire mixed sand shows the optimum results for the construction of MSE wall. By using rubber in wall, we can reduce the harmful effect produced due to dumping and burning of tires.
Keywords: Crumb tire rubber; MSE wall; Soil reinforcement; Shear strength; Direct shear box test
Paper 5: PERFORMANCE ANALYSIS OF OFF THE SHELF AND CUSTOMIZED TWITTER SENTIMENT CLASSIFICATION MODELS (27-32)
Abstract: : : Advent of social networks has provided voice to common people. People utilize these medium to raise their issues or express their point of view about a trending topic. Especially Twitter, that has become de facto standard to discuss current issues, problems, events and policies. This scenario demands computer mediated systems that may be able to produce their collective opinion. Research is being carried out to develop sentiment classification models that may use social media data through general purpose programming languages like python or by using off the shelf tools such as Rapid Miner. This work is an attempt to know the comparative suitability of customized built models using python and off the shelf tools made models using Rapid Minder on twitter text data. We found customized built sentiment classification models more efficient, flexible and robust than off the shelf solutions. Customized twitter sentiment classification model, developed using python, achieved 3-4 points more accuracy than off the shelf classification models. Additionaly custom sentiment classification models takes less amount of time in building dictionary, training models and predicting class labels for given tweets.
Keywords: Machine learning; Off the shelf tools; Rapid Miner; Sentiment analysis; Twitter Sentiment Classification (TSC)
Paper 6:VOICE TO TEXT SYSTEM-EMMA A DESKTOP APPLICATION (33-35)
Abstract: In this paper we present smart and intelligent computer-based application. This application has been designed to provide efficient and fast ways of searching and performing the operation whatever is required. The proposed application will be communicating to its 455555 user and also the user will be able to communicate with it with the help of voice commands as well as by text commands also. This desktop application is named as EMMA. This is an Artificial Intelligence based which is able to respond to the requirements of the users and based on their requirements it will perform operations.
Keywords: Voice, Message, Desktop, Intelligent
ISSN: (e) 2520-7393
(P) 2521-5027
Indexing and abstracting services
