A filter-based feature selection approach for the prediction of Alzheimer's diseases through audio classification
Vikas Yadav, Rahul Kumar, Chandrashekhar Azad · 2022 2nd International Conference on Advance Computing and Innovative Technologies in Engineering (ICACITE) · 2022
Alzheimer's disease (AD) is a accelerating brain disease that wreaks havoc on memory and cognition, as well as the ability to perform even the most basic tasks. Dementia, which gradually damages brain cells, is the primary cause of AD. People with this disease lose their ability to think, read, and do a variety of other things. A machine learning system can help to alleviate this problem by predicting the disease. The primary goal is to detect dementia in a wide range of patients. The classification of speech signal into various categories such as audio, non- audio is an very significant front-end problem in AD prediction. There are many features have been presented for AD prediction. Unfortunately, these characteristics are not mutually exclusive, and combining them has no effect on AD prediction performance. By removing these redundant and irrelevant features, Feature Selection (FS) provides a simple yet effective solution to this problem. Removing irrelevant data improves learning accuracy, decreases computation time, and allows for a more thorough understanding of the learning model or data. The proposed model in this paper is a hybrid machine learning model in which we used an FS approach based on Mutual Information (MI). MLP is used for classification and MFCC is used to extract all of the features from the ADReSSo dataset. In terms of accuracy, our experimental results show that the proposed classifier outperforms existing classifiers.