نوع مقاله : علمی-پژوهشی
نویسندگان
1 گروه مهندسی پزشکی، دانشکده مهندسی پزشکی، دانشگاه صنعتی سهند، تبریز، ایران
2 گروه مهندسی بین رشتهای، دانشکده علوم مهندسی، دانشکدگان فنی، دانشگاه تهران، تهران، ایران
چکیده
کلیدواژهها
موضوعات
عنوان مقاله [English]
نویسندگان [English]
Since the content of dreams can mirror an individual's mental state, recognizing emotions during sleep can reveal valuable information about the state of mind. This research suggests a new method for the automated classification of emotions during the state of sleep known as the rapid eye movement (REM) state. In the suggested method, in the preprocessing step, electroencephalogram (EEG) signals are filtered with a Butterworth bandpass filter and are then broken into 20-second windows. Each filtered EEG window is then decomposed into five intrinsic mode functions (IMFs) with the empirical mode decomposition algorithm. In the next step, the sample entropy feature is extracted from each IMF. The ReliefF algorithm is utilized to select the best subset of features, which are applied as an input to random forest, support vector machine, and K-nearest neighbor machine learning algorithms. Emotional states are classified in a three-class setting (positive, negative, and neutral) as well as in two binary classification scenarios: (1) positive versus negative, and (2) neutral versus a combination of positive and negative emotions. EEG signals from the publicly available dataset known as the DEED dataset are utilized to see brain activity patterns which are present in different states of emotions in dreams and to analyze the performance of the suggested method. The outcome reveals that the suggested method provides an accuracy of 93.74% in three-class classification and an accuracy of 97.45% in binary classification (emotions versus neutral), which indicates good performance in the classification of different states of dreams' emotions during sleep.
کلیدواژهها [English]
)