Emotion Recognition from DEAP Dataset Using SVM Classifier
Sachin1, Deepak Kumar2
1 Department of Electrical Engineering, Om institute of technology and management, Hisar- 125001, Haryana, India; sachin.svk93@gmail.com.
2 Department of Electrical Engineering, Om institute of technology and management, Hisar- 125001, Haryana, India; ergargdeepak@gmail.com.
Abstract: Emotion recognition has become critical for facilitating and enhancing human-computer interaction. Despite the critical role emotions play in human communication, the majority of existing human-computer interaction systems are incapable of recognising and interpreting user emotions. Internal physiological signs such as heart rate, breathing, and brain activity may be used to detect emotions. Physiological cues, particularly brain impulses, are regarded more trustworthy for emotion detection. Numerous academics and businesses have shown an interest in deciphering user intents through brain signals during the past few years. The most often utilised method for monitoring brain activity is electroencephalography (EEG). The EEG monitors the brain's electrical activity using electrodes implanted on the scalp. It provides great temporal resolution without posing any hazards, and it is very inexpensive. Several commercial EEG devices have been developed over the past several decades, and these devices are much simpler to set up and operate than laboratory-based EEG equipment.
Recognizing emotions from brain signals is not a simple job, since emotion representation is complex and gender dependent. The majority of prior research had poor accuracy in recognizing emotions, and some of them created a model for each user individually or for a subset of users. The purpose of this study is to enhance emotion identification by analyzing brain signals collected by EEG equipment and classifying male and female emotions. To decrease noise in EEG signals, many stages must be performed: reference, segmentation, band pass filtering, and denoising using wavelet transform. From the noise-free data, EEG frequency bands are isolated and five characteristics are computed to provide the feature vector for the classifier. Finally, we utilise an SVM classifier to identify the user emotion associated with the retrieved feature vector. The experimental findings established the suggested model's superiority and resilience in comparison to previous research that utilised the same dataset. The suggested model demonstrated a greater degree of accuracy (99.02%) when compared to the findings of previous research.
Keyword: Classifier, SVM, Machine Learning, EEG.
