Detecting Mild Cognitive Impairment Through Voice Elements and Gait Factors from the Dual Task of Mental Arithmetic and Walking

Recently, it has been reported that the performance of dual tasks (the execution of two tasks simultaneously) could enable the detection of mild cognitive impairment. The authors believe that task execution affects the vocal cords. It can be inferred from this premise that the effect on vocal cords would be greater during dual-task execution than during single-task execution. Therefore, we sampled the voices of healthy elderly subjects and elderly subjects who showed indications of mild cognitive impairment during single- and dual-task execution, created estimation models of the cognitive ability test score from the prosodic features of the voices, and compared the estimation accuracy of these models. The single task was an arithmetic exercise in which the subject was required to subtract a series of 7 from 100 (the serial 7 task). The dual task constituted walking a fixed distance while performing the serial 7 task. The results of accuracy testing on data other than the training data showed a weak positive correlation between the actual cognitive ability test scores and their estimates in the model created from the voice features from the dual task. A weak positive correlation was also obtained between the actual cognitive ability test scores and their estimates for the model created from the gait factors from the dual task for data other than training data. Thus, the results suggest that voice elements and gait factors from the dual task can be used effectively for the detection of mild cognitive impairment.