Scholarlink

<p>The most prevalent sense of impairment in the world, hearing loss hinders learning and communication. The best way to address this issue is to use electroencephalograms (EEGs) to detect hearing loss early and accurately. The most relevant modality for hearing loss among the several EEG control signals is the auditory evoked potential (AEP), which is generated by the brain in response to an auditory input. Consequently, a new approach is suggested in this study to ascertain the degree of hearing sensitivity based on the AEP response. In this paper, a Fused Perceptron with Chi Squared Multi-Kernel based Extreme Learning (FPCSMK-EL) method is presented for more accurate detection of hearing sensitivity levels. Preprocessing, feature extraction, and classification are the three distinct processes carried out by the suggested FPCSMK-EL approach. Initially, the Fusion procedure known as Fast Independent Component Analysis (FICA) and Empirical Mode Decomposition (EMD) are used to preprocess the EEG signal in order to minimize the error rate during the identification process by eliminating artifacts. After that, Haar Wavelet Multilayer Perceptron and Kullback–Leibler Divergent Chi-Square Model are applied to extract the relevant EEG features from AEP. Here, Haar Wavelet Multilayer Perceptron is employed to determine the threshold factor for hearing (i.e., parameter modeling) and Kullback–Leibler Divergent Chi-Square Model is used to extract the features (i.e., non-parameter modeling) with minimum time. Through the extraction of relevant EEG features, overhead incurred in the hearing sensitivity level detection is reduced. Lastly, Multi-kernel extreme learning machine classifier model is applied to perform classification based on the relevant features of EEG signal for detecting hearing sensitivity level with higher accuracy. A MATLAB simulation tool with a variety of performance measures is used to analyze the proposed FPCSMK-EL method. The proposed FPCSMK-EL method has an average recall of 95.55%, a detection overhead of 2.624&#xa0;KB, and detection error rate of 3.445% are surpasses the existing LSTM and AAD-transformer techniques.</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

Fused Perceptron and Multi-kernel Extreme Learning for Hearing Sensitivity Level Detection with Auditory Evoked Potentials

  • 期刊论文

  • 发表日期：

    2025年6月19日

查看全文
  • K. M. Kirthika,
  • M. Paulraj,
  • N. Sanjana,
  • R. Immanual

摘要

The most prevalent sense of impairment in the world, hearing loss hinders learning and communication. The best way to address this issue is to use electroencephalograms (EEGs) to detect hearing loss early and accurately. The most relevant modality for hearing loss among the several EEG control signals is the auditory evoked potential (AEP), which is generated by the brain in response to an auditory input. Consequently, a new approach is suggested in this study to ascertain the degree of hearing sensitivity based on the AEP response. In this paper, a Fused Perceptron with Chi Squared Multi-Kernel based Extreme Learning (FPCSMK-EL) method is presented for more accurate detection of hearing sensitivity levels. Preprocessing, feature extraction, and classification are the three distinct processes carried out by the suggested FPCSMK-EL approach. Initially, the Fusion procedure known as Fast Independent Component Analysis (FICA) and Empirical Mode Decomposition (EMD) are used to preprocess the EEG signal in order to minimize the error rate during the identification process by eliminating artifacts. After that, Haar Wavelet Multilayer Perceptron and Kullback–Leibler Divergent Chi-Square Model are applied to extract the relevant EEG features from AEP. Here, Haar Wavelet Multilayer Perceptron is employed to determine the threshold factor for hearing (i.e., parameter modeling) and Kullback–Leibler Divergent Chi-Square Model is used to extract the features (i.e., non-parameter modeling) with minimum time. Through the extraction of relevant EEG features, overhead incurred in the hearing sensitivity level detection is reduced. Lastly, Multi-kernel extreme learning machine classifier model is applied to perform classification based on the relevant features of EEG signal for detecting hearing sensitivity level with higher accuracy. A MATLAB simulation tool with a variety of performance measures is used to analyze the proposed FPCSMK-EL method. The proposed FPCSMK-EL method has an average recall of 95.55%, a detection overhead of 2.624 KB, and detection error rate of 3.445% are surpasses the existing LSTM and AAD-transformer techniques.