Modes of Operation and Approaches

Two are the modes of operation for the software, namely Gait and Training mode:

Gait Mode

In Gait mode, the characteristics of the patient's walking are detected, with auditory feedback being provided upon detection of any irregularities. Data is classified by means of a convolutional neural network (CNN) into three classes, namely healthy, hemiplegic, and foot drop gaits. A four-layered CNN was extensively trained using 4 healthy subjects, who were required to walk at slow, average, and fast pace for several minutes, and through simulations of the hemiplegic and foot drop gaits. Acceleration and gyro data from the 3 IMUs mounted on the affected limb, and from the pressure sensors embedded in both insoles were fed in the CNN for this purpose. In order to train the CNN, data was processed offline, using a bandpass filter, followed by appropriate chunking and resampling.

Training Mode

Training mode offers typical game-based rehabilitation exercises, including weight shifting and calf pushes, which are streamed to an external device. During weight shifting, the user is required to move his/her weight from one foot to the other to reach the force targets set in two display bars, corresponding to fractions of the body weight, which was pre-recorded during calibration. Similarly, targets are set in the calf pushes exercises, whereby the user is to shift his/her weight from heel to toes. In this exercise, targets are displayed in the form of concentric rings.

Click here to go back to the Software section of the Wiki.