RETRACTED ARTICLE: Automated speech based evaluation of mild cognitive impairment and Alzheimer’s disease detection using with deep belief network model
Chiai AI-Atroshi, J. Rene Beulah, Kranthi Kumar Singamaneni, C. Pretty Diana Cyril, S. Neelakandan, S. Velmurugan
Abstract
Early detection of Moderate Cognitive Impairment (MCI) and Alzheimer's disease (AD) is critical for increasing survival rates. Speech feature extraction is used in prior MCI and AD detection algorithms during neuropsychological assessments by medical specialists. The study's goal is to create an MCI and AD detection model using Automatic Speech Recognition (ASR) and a DL model. The suggested approach largely employs the Gaussian Mixture Model (GMM) for ASR in the patient's spontaneous speech. Furthermore, the Deep Belief Network (DBN) model is used to extract feature vectors from identified voice data. Finally, the SoftMax (SM) classifier is used to detect the presence of MCI and AD disorders in the used speech signals. A series of simulations were run to assess the superior performance of the GMM-DBN (Gaussian Mixture Model-Deep Belief Network model). Effectively describes the distribution of data observations as a weighted average of parameterized Gaussian distributions. The testing results indicated the GMM-DBN model's superior performance, with maximum accuracy of 90.28% and 86.76% on the binary and multiple class classifications, DN respectively. The GMM-DBN methodology has been successful in the classification of multiple classes, as evidenced by its F1-score reaching a maximum of 90.19% and its accuracy reaching 90.28%.