Abstract:

Objective To improve the anti-fall capacity and safety of the smart walkers.

Methods Two pressure sensors were placed on the handles on both sides of the walker. The confusion matrix was obtained, the corresponding operational intent labels were manually labeled, using a support vector machine (SVM) classifier for model prediction to predict the travel intent of the users. The user wore a gyroscope and the walker was equipped with a laser sensor, to measure the angular velocity, angular acceleration and the distance data, respectively, to detect the user's fall.

Results The classifier model established by SVM successfully predicted three operating states of the walker, namely straight ahead, left turning and right turning. The user's fall was detected by the sudden change of the following data: the combined angular velocity was greater than 100°/s, the combined angular acceleration was greater than 1.3 G, the angular acceleration of Z-axis was greater than 0.7 G or less than 0.2 G, and the distance was greater than 600 mm or less than 300 mm.

Conclusion The improvement of the walker can predict the turn intention of the user, and detect the user's fall.

Key words: walking aid robot, turn intention perception, fall detection