Publications
Human Facial Aggressive Detection System Based on Facial-Width-to-Height Ratio
Sep 10, 2020Journal IOP Conference Series: Materials Science and Engineering
Publisher IOP
DOI 10.1088/1757-899X/978/1/012046
Volume 978
Recent researches have shown closely related evidence between the human individual social behavior and precisely measurable facial features. The Facial-width-to-height ratio (FWHR) has become quite an interesting topic concerning human aggressive behavior. Recent studies presented evidence showing that the precise measurement of FWHR can be used to predict human aggressive behavior based on facial landmark extraction. In this paper, the Facial-width-to-height ratio is extracted and analyzed among men, women, and children using the recently presented Convolutional Experts Constrained Local Model (CE-CLM). Then, extracted features are used to train the Numeral Virtual Generalizing Random Access Memory (NVG-RAM) pattern recognition technique. The results show promising clues in depending on this feature extraction method for the Facial-width-to-height ratio, and depending on SVG-RAM classifier for aggressive behavior. Moreover, the proposed method is less susceptible to facial rotation error ensuring accurate FWHR extraction. © Published under licence by IOP Publishing Ltd.
FPGA Design and Hardware Implementation of Heart Disease Diagnosis System Based on NVG-RAM Classifier
Apr 9, 2019Publisher IEEE
DOI 10.1109/SCEE.2018.8684125
This paper presents a diagnosis system design used to assist the physicians to diagnose the heart condition by converting medical factors of the patients into a numerical representation. The proposed heart disease diagnosis system can classify two heart conditions (normal and abnormal). Also, it can classify four abnormality heart conditions in addition to the normal case. Two types of database are used in the classification process: the online database from The University of California, Irvine (UCI) machine learning dataset repository and collected real database (CD). These databases consist of 13 medical factors that are successful in diagnosing heart disease. The simulation results show that, the proposed Numeral Virtual Generalizing Random Access Memory (NVG-RAM) Weightless Neural Network classifier has 100% accuracy of two heart diseases classification when the performance of this classifier was evaluated using CD. Additionally, this classifier achieves 90% success rate when recognizing 5 states for the same database. According to the UCI database the NVG-RAM is considered best classifier for classifying two types of heart disease based on different division of training and testing database. Furthermore, the diagnosis accuracy for classifying five types is 71.698%. The proposed Heart disease classifier is hardware implemented using FPGA platform kit (Spartan-3A DSP 3400A). This classifier achieves high success rate when tested in using CD for diagnosis two-class heart disease that gives maximum accuracy 100%. Moreover, the NVG-RAM is considered a good algorithm to diagnosis multiclass heart diseases that gives a maximum accuracy of 88%.