Abstract:
The emotional wellness of a person includes one’s thoughts, emotions, and ability to deal with life’s challenges. A sign of emotional wellness is having the ability to talk with someone about your emotional concerns and share your feelings with others. A psychotic patient is not emotionally well, and it is very important to know to what extent the patient is unwell. To answer this question, this research investigated the negative emotions of humans. The researchers have used electro Enspheloraphy (EEG) method to find emotions, however, in this study, the use of an emerging technology of the present, called fNIRS (functional near infra-red spectroscopy), is selected because the accuracy of correct detection of emotions using this technology is proved by researchers. However, the detection of negative emotions has not been studied yet. In this research data from 10 healthy patients is recorded for three emotions: sad, angry, and neutral. To record emotions the fNIRS data acquisition system was used. The data is filtered, features are extracted and then data is classified for the three emotions mentioned above. One deep learning model and five machine learning models named the LSTM (Long Short Term Memory), Decision Tree, K-NN (K-Nearest Neighbours), Random Forest, SVM (Support Vector Machine), and Naive Bayes are applied to classify the data for three emotions. The percentage maximum accuracy of the emotion detection out of these models is 99% through LSTM, whereas, a minimum accuracy of 76% is achieved through Naive Bayes. The remaining models gave accuracies of 98% through the Random forest, the Decision tree, and the K-NN, and 88% through SVM. The accuracies are found improved as compared to those achieved through EEG as per existing literature. This verifies the efficacy of the methodology that acquires data through fNIRS technology and is classified using different classifiers.
