Emotion recognition through electroencephalography (EEG) signals is an important aspect of human-computer interaction that poses a significant research challenge. Most of the current approaches utilize up to 18 channels from 32 available channels for extracting emotions features. Moreover, they only use valence and arousal model to classify emotions. Therefore, the current approaches are unable to detect emotions accurately. In this paper, a framework that utilizes a three-dimensional model incorporating arousal, valence, and dominance for identifying emotions is proposed. Our framework can define any number of emotions, even in the absence of discrete emotions labels. The electroencephalography signals from DEAP database are utilized for emotion detection. The effectiveness of three classification techniques is examined, namely Support Vector Machine (SVM), K-Nearest Neighbors (KNN), and Multilayer Perceptron (MLP). The classification accuracies for valence, arousal, and dominance are 90.19%, 91.91%, and 89.86%, respectively. Our results demonstrate that EEG data's time-domain statistical and power features can effectively classify different emotional states. Furthermore, our framework enables accurate identification of identical emotions that cannot be distinguished by a two-dimensional model.