Detection of Intra-Personal Development of Cognitive Impairment From Conversational Speech

As the population in developed countries is aging, cognitive impairment such as Alzheimer's disease becomes an urging challenge for these societies. In order to mitigate the consequences, diagnosing cognitive impairment early is crucial. We present automatic detection of an intra-personal development of cognitive impairment from speech. Using conversational speech data from the ILSE corpus we detect subjects which were considered cognitively healthy at one examination and were diagnosed with a cognitive impairment at a later examination. From the speech recordings we extract 14 speech-based features using voice activity detection and transcriptions. With these features we train a linear discriminant analysis classifier that distinguishes subjects who developed a cognitive impairment from subjects who did not. The classifier achieves an accuracy of 80.4%, classifying half the cognitively impaired subjects correctly and assigning that label to hardly any cognitively health subjects. This shows our approach is well suited for longitudinal cognitive status monitoring.