Scholarlink

<p>In recent years, emotion recognition and classification has gained significant attention in fields like human-computer interaction and mental health. Emotion recognition aim to recognize and classify emotions of a human based on various data sources like facial expression, body language, voice, and text. Many state-of-the-art demonstrated the use of imagery data and electroencephalography signals to conduct automatic emotion recognition. These systems are often struggling with accuracy in recognizing emotions and not generalized well to real-world scenarios, leading to low performance in practical applications. This paper presents a new emotion recognition and classification framework, which utilizes electroencephalography signals as data source. The goal of this paper is to develop a framework that accurately recognizes and classifies three classes of human emotions like positive, negative, and neutral from a SEED dataset. The proposed framework is a lightweight, consists of three phases like data gathering and preprocessing, feature extraction and selection, model development and evaluation. During the classification process three machine learning classifiers like random forest, support vector machine, and neural network classifier are utilized. The proposed framework is trained on original dataset and the augmented dataset to improve the classification performance. During the process, the evaluation is conducted using different metrics like overall accuracy, precision, recall, f1-score, and Cohen’s kappa, and confusion matrices. The proposed framework achieves nearly 81% of accuracy in classifying the emotion types with random forest classifier that is 4.76% higher than support vector machine and neural network classifiers. The data augmentation technique has been applied to improve the classification accuracy. The Random Forest classifier has achieved 91% accuracy on augmented dataset that s 10% higher than the original dataset. The proposed work will advance the need of emotion recognition systems for building applications that can significantly enhance the human-computer interaction and provide valuable insights into human emotions.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Electroencephalography-based emotion recognition and classification using machine learning algorithms

  • 期刊论文

  • 发表日期：

    2026年2月26日

查看全文
  • K. Indraneel,
  • Trinath Basu Miriyala

摘要

In recent years, emotion recognition and classification has gained significant attention in fields like human-computer interaction and mental health. Emotion recognition aim to recognize and classify emotions of a human based on various data sources like facial expression, body language, voice, and text. Many state-of-the-art demonstrated the use of imagery data and electroencephalography signals to conduct automatic emotion recognition. These systems are often struggling with accuracy in recognizing emotions and not generalized well to real-world scenarios, leading to low performance in practical applications. This paper presents a new emotion recognition and classification framework, which utilizes electroencephalography signals as data source. The goal of this paper is to develop a framework that accurately recognizes and classifies three classes of human emotions like positive, negative, and neutral from a SEED dataset. The proposed framework is a lightweight, consists of three phases like data gathering and preprocessing, feature extraction and selection, model development and evaluation. During the classification process three machine learning classifiers like random forest, support vector machine, and neural network classifier are utilized. The proposed framework is trained on original dataset and the augmented dataset to improve the classification performance. During the process, the evaluation is conducted using different metrics like overall accuracy, precision, recall, f1-score, and Cohen’s kappa, and confusion matrices. The proposed framework achieves nearly 81% of accuracy in classifying the emotion types with random forest classifier that is 4.76% higher than support vector machine and neural network classifiers. The data augmentation technique has been applied to improve the classification accuracy. The Random Forest classifier has achieved 91% accuracy on augmented dataset that s 10% higher than the original dataset. The proposed work will advance the need of emotion recognition systems for building applications that can significantly enhance the human-computer interaction and provide valuable insights into human emotions.