计算机与数字工程2025，Vol.53Issue(4)：1081-1085,1200,6.DOI:10.3969/j.issn.1672-9722.2025.04.028
基于K近邻的情绪识别中的噪声效益
Noise Benefit in Emotion Recognition Based on K-nearest Neighbor
摘要
Abstract
Stochastic resonance(SR)theory studies the synergistic effect of noise and weak input signals in nonlinear sys-tems,showing that noise of appropriate intensity can improve system performance.Based on the stochastic resonance theory,this pa-per discusses the noise benefit in the process of emotion recognition using EEG signals.Firstly,independent Gaussian noise is add-ed to the original signal and wavelet packet transform(WPT)is performed on it.Secondly,the energy characteristics of the effective frequency bands are extracted and normalized.Finally,it is recognized by K-nearest neighbor(KNN)classifier and the classifica-tion accuracy is calculated.Through the experiment of binary classification(calm/pressure state)on DEAP dataset,the results show that the average classification accuracy of the system can be effectively improved by adding appropriate intensity Gaussian noise at different k values.When k=5,the average classification accuracy of the system after adding noise is increased by 17.78%compared with that without adding noise.When k=15,the maximum average classification accuracy of the system with noise is 78.33%,which is 4.26%higher than that without adding noise,which shows the effectiveness of this method.关键词
随机共振/高斯噪声/情绪识别/K近邻Key words
stochastic resonance/gaussian noise/emotion recognition/K-nearest neighbor分类
通用工业技术引用本文复制引用
刘鑫,翟其清,王友国..基于K近邻的情绪识别中的噪声效益[J].计算机与数字工程,2025,53(4):1081-1085,1200,6.
