Detection of Alzheimer’s Disease Through Speech Features and Machine Learning Classifiers

Abstract

AbstractThe most common neurodegenerative disease identified in elderly people is Alzheimer’s disease (AD). In general, identification of Alzheimer’s is carried out based on medical images of the brain. The other non-invasive signal processing techniques for AD detection are reviewed. Possibly, these reviewed techniques can also provide an early diagnosis of AD, which might help delay the development of the disease. These models are based on the spontaneous speech (SS) features extracted, such as vocal, linguistic, acoustic, prosodic, and features. Various classifiers and machine learning algorithms that are feasible in AD detection are described. These are discussed and compared based on their performance accuracies of classification results. Few models developed on repository corpuses and others based on speech data collected from persons of different age groups, affected with AD, having mild cognitive impairment (MCI) and/or healthy control (HC) are presented. The models reviewed may help further research toward the development of a reliable machine assistive technology (MAT) for providing health care to elderly people and exploring cures or solutions to the diseases like AD, PD, vascular dementia, down syndrome, frontotemporal dementia, etc.KeywordsSpeech recognitionAlzheimer’s disease (AD)Spontaneous speech (SS)Mild cognitive impairment (MCI)Healthy control (HC)Machine assistive technology (MAT)