Wearable Sensors Based Automatic Box and Block Test System

摘要

Objective manual dexterity assessment is useful for early detection of age-related functional decline and for prediction of cognitive decline. Such assessments are carried out by clinicians using traditional tests and scales. The Box and Blocks Test (BBT) is one of such scale, focusing on the measurement of unilateral gross manual dexterity. Subjects are scored based on the number of blocks transferred from one compartment to the other compartment in a time window. Currently, most works on automatic BBT take advantage of the data captured by deep camera. However, these kind of solutions are limited by computational complexity and the accuracy of color segmentation. Based on recent success of wearable equipments, in this paper, we research multimodal wearable sensors based Box and Block Test system (wBBT). wBBT takes advantage of 5 kinds of time series signal, including Electromyographic, Accelerometer, Gyroscope, Orientation and Orientation Euler for automatic manual dexterity assessment. In addition, we research 9 deep neural network structures (3 multimodal signal fusion methods × 3 network cells) for multimodal signal based blocks moving motion recognition. To evaluate the feasibility of wBBT, we collect the data of 37 subjects, including 25 subjects with Mild Cognitive Impairment (MCI) and 12 normal subjects. Experimental results show that wBBT is a useful tool for automatic assessment of unilateral gross manual dexterity.