Classifying Alzheimer's Disease Using Audio and Text-Based Representations of Speech

摘要: Alzheimer's Disease (AD) is a form of dementia that affects the memory, cognition, and motor skills patients. Extensive research has been done to develop accessible, cost-effective, non-invasive techniques for automatic detection AD. Previous shown speech can be used distinguish between healthy patients afflicted In this paper, ADReSS dataset, dataset balanced by gender age, was automatically classify AD from spontaneous speech. The performance five classifiers, as well convolutional neural network long short-term memory network, compared when trained on audio features (i-vectors x-vectors) text (word vectors, BERT embeddings, LIWC features, CLAN features). same were train regression models predict Mini-Mental State Examination score each patient, maximum value 30. top-performing classification support vector machine random forest classifiers which both achieved an accuracy 85.4% test set. best-performing model gradient boosting embeddings had root mean squared error 4.56 tasks illustrates feasibility using neuropsychological scores.