•  
  •  
 

Abstract

Waveform detection has been an area of continuous investigation for many years. One important waveform in sleep stage 2 is the k-complex. Numerous researchers have created various strategies for auto-k-complex detection; some of these strategies state that the automated detection techniques are adequate. Because of its analytically relevant resolution, the Electroencephalogram (EEG) is a commonly utilized technique to analyze the k-complexes in order to understand the nervous system activity of the brain. Several researchers have classified waveforms using EEGs in a variety of ways. It appears that the majority of the waveform detection had limitations. The necessary analyzes took a lot of time to complete, no execution time was indicated in any of the earlier research, and they were too complex for real-world use. Furthermore, it was revealed that numerous experiments were done without window size and were utilized to detect waveform characteristics. Additionally, the research used one or two evaluation instruments to analyze the performance outcomes. For the dataset, a maximum accuracy of 94% to 75% was reported. Because of its significance, several analysts have developed an automated technique to use EEG data to study k-complexes. This work proposes a novel approach to feature detection for k-complexes utilizing a least square support vector machine (LS-SVM) classifier. The sliding window method divides EEG signals into a number of segments. Subsequently, distinct feature sets are obtained from every time interval. Every EEG segment was visible in the obtained twenty-seven features as vectors. That means, twenty-seven features were extracted using the Katz algorithm and the Tunable Q-factor wavelet transform for each segment. These features were analyzed, to select the most important features, using an Analysis of Variance (ANOVA) and the F-test. Finally, the vector of features was used as input to the LS-SVM classifier. When it came to identifying events of (non) k-complexes, the suggested novel technique demonstrated noteworthy performance results with sensitivity, accuracy, and specificity of 98.3%, 96.5%, and 91.6%, respectively. This high accuracy rate has not been discovered in any method yet. When compared to other classifiers and methods in this field of research, the LS-SVM classifier approach yielded the best results.

Creative Commons License

Creative Commons License
This work is licensed under a Creative Commons Attribution-Noncommercial-No Derivative Works 4.0 License.

Share

COinS