首页|Reports from Indira Gandhi Delhi Technical University for Women Add New Data to Findings in Machine Learning (Novel Fractal Pattern Based Features for Eeg-based Emotion Identification)
Reports from Indira Gandhi Delhi Technical University for Women Add New Data to Findings in Machine Learning (Novel Fractal Pattern Based Features for Eeg-based Emotion Identification)
扫码查看
点击上方二维码区域，可以放大扫码查看
原文链接
NETL
NSTL
Investigators publish new report on Ma chine Learning. According to news reporting out of Delhi, India, by NewsRx edito rs, research stated, "Comprehending the manifestation of emotional states in EEG signals is a well -established field of research spanning neuropsychology and b iomedical signal processing. Two main approaches exist for classifying emotions based on EEG data: (1) extracting features and employing machine learning algori thms, and (2) deep learning." Our news journalists obtained a quote from the research from Indira Gandhi Delhi Technical University for Women, "Machine learning faces the major challenge of poor accuracy due to low inter -class variability and low intra-class similarity and deep learning is challenged by the high computational cost. The authors hav e addressed the above challenges by devising a novel fractal based EEG feature e xtraction method so as to achieve better performance using machine learning thus avoiding high computational complexity (similar to deep learning). Fractal patt erns being capable of providing a measure of self similarity at different scales , have been used to capture EEG signal's intricate dynamics. These patterns bein g resistant to noise and artifacts, provide a better reflection of the underlyin g organization and complexity of brain activity caused due to different emotions . The above listed inherent capabilities make fractal pattern suitable for provi ng valuable insights into cognitive processes and neurological disorders. So, th e authors have explored fractal pattern for local feature generation from Tunabl e -Q Wavelet Transform (TQWT) features. Subsequently, a minimum redundancy and m aximum relevance (mRMR) selector is applied to select the most informative featu res. These selected features are then fed to established classification algorith ms for emotion recognition. The proposed framework is evaluated on two publicly available datasets, DEAP and DREAMER, achieving impressive accuracy rates of 95. 43% and 99.14%, respectively."
DelhiIndiaAsiaCyborgsEmerging Te chnologiesMachine LearningIndira Gandhi Delhi Technical University for Women
NETL
NSTL
