hilbert-mel frequency spectrum features for efficient eeg-based alzheimer's detection

Alzheimer's disease (ad) is a progressive neurodegenerative disorder that severely impairs cognitive function and disrupts brain connectivity. early and accurate diagnosis is crucial for effective intervention, yet identifying discriminative features from complex electroencephalography (eeg) signals remains a challenge. resting-state eeg provides a non-invasive and cost-effective tool for ad detection, but its diagnostic utility is highly dependent on the quality of extracted features. this study introduces a novel feature extraction approach that uses mel-frequency spectrum features (mfs) and the hilbert transform (ht) to enhance both spectral and temporal feature representation of eeg signals. the proposed hilbert-mel frequency spectrum (hmfs) framework captures subtle variations in phase and amplitude, providing a rich and complementary set of descriptors. principal component analysis (pca) is employed to reduce dimensionality while retaining key information, enabling more efficient and accurate classification. a 5-fold cross-validation approach was employed to assess model performance and generalizability. the extracted features are classified using various machine learning models, with k-nearest neighbors (knn) achieving the highest performance. the proposed method reached an accuracy of 99.24% with a perfect recall of 100%, precision of 98.61%, specificity of 98.39%, f1-score of 99.30%, and geometric mean score of 99.31%. compared to existing eeg-based ad detection techniques, the hmfs method surpasses previous approaches in accuracy and recall and the method achieves higher performance. the integration of spectral and temporal features results in a more robust feature space, thereby improving generalization. this approach provides a reliable, efficient framework for early ad diagnosis with potential clinical applications.