الفهرس 
IntroductionOnly 14 pages are availabe for public view
 |  | from | 244 |  |  |
| | | from | 244 | | |
Abstract
Emotion identification is the first step towards emotional intelligence in advanced human-machine interaction. AHER which stands for (Automated Human Emotion Recognition) is a critical topic investigated by computer science. It can be used in many applications such as marketing, human-robot interaction, electronic games, E-learning, and others. Automatic recognition of human emotions is not a trivial process. There are many factors affecting emotions internally and externally. Expressing emotions could also be done in many ways such as text, speech, body gestures, or even physiologically by physiological body responses. The availability of advanced technologies such as mobiles, sensors, and data analytics tools, led to the ability to collect data from various sources, which enabled researchers to accurately predict human emotions. Most current research uses them in lab experiments for data collection. Because of urbanization or technological advance, energy use, trash creation, and other variables have no impact. The urban environment can have a significant effect on our bodies. Based on that, emotion identification using physiological signals and environmental variables has lately emerged.
Goals: -
1) A review and an evaluation of the state-of-the-art methods for AHER employing Machine Learning from a computer Science perspective.
2) Combining both on-body physiological markers, surrounding sensory data, and emotional measurements to achieve the following goals:
• Collecting a multi-modal data set including environmental factors, body responses, and emotions in real settings to predict emotional reactions.
• Studying the impact of environmental factors on the human body responses such as Heart rate HR, Electro-dermal Activity (EDA), Body temperature (btemp), and motion using multiple linear regression analysis.
• Creating Predictive models of emotional states based on various modalities associated with several homogeneous sensors using various learning methods.
• Assessing ensemble learning methods and comparing their performance for creating a generic subject-independent model for emotion recognition with high accuracy.
• Providing a framework (System architecture) for fusion-based multi-modal emotion recognition.
• Implementing various ML Methods to find the best efficient ways and find the best algorithm to classify five distinct emotional states ranging from very negative to very positive to construct a user-dependent model based on fusing environmental and on-body variables.
Results: -
- First, environmental factors have variant effects on the human body. Some environmental variables have significant effects on body responses such as air pressure and ambient light levels UV. Surprisingly, environmental noise has little effect on various body responses.
- Second, results proved that RF and DT are the best base algorithms and outperformed SVM, and KNN with an average accuracy (of 97%), where SVM predicts an average accuracy for the emotion of (95%) and KNN with an average accuracy of (94%) for all participants to create a user-dependent or personalized emotion model using four ML algorithms.
- Third, the ensemble stacking learner technique gave the best accuracy of 98% compared with other ensemble learning methods. On the contrary, bagging and boosting methods gave (96.4%) and (96.6%) accuracy levels to build a user-independent emotion model based on integrating on-body and ambient sensors using ensemble methods.
- Finally, to avoid any over-fitting or under-fitting, grid-search hyper-parameter, cross-Validation, and Random Train-Test Split were utilized to get an accurate result and find the best hyper-parameters for each model.