Skip to main content
Fig. 5 | Journal of NeuroEngineering and Rehabilitation

Fig. 5

From: Auto detection and segmentation of daily living activities during a Timed Up and Go task in people with Parkinson’s disease using multiple inertial sensors

Fig. 5

Comparison between the original and modified algorithms during walking. The original algorithms used to detect walking were based on the gyroscope of the sacrum (ωy) and the modified algorithms were based on the acceleration of the shin (a y) and hip angle (θ hip). The original algorithms were prone to false positive during sitting down (a). Furthermore, due to minimal hip movement in the y-direction in some participants who exhibited shuffling gait, detection was more prone to false negative since the signal dropped below the normalized threshold. To remedy this problem, an adaptive threshold was used to set the limit of detection based on the distribution of the amplitude of the signal. This approach adaptively changes the threshold based on the participants; therefore, mitigating the variability among participants (b). The acceleration of the shin was adopted to detect walking in PD patients; however, using this IMU alone also yielded many false positives due to extraneous lower limb movements during sitting down and standing up. To identify true moment of walking, θ hip was used to ensure that the participant was standing upright. Therefore, the movement of the shin coupled the upright position of the participant distinguished walking from other tasks with 100% accuracy (n = 72) (c)

Back to article page