We present an automated gait segmentation method based on the analysis of foot plantar pressure patterns elaborated from two wireless pressure-sensitive insoles. The 64 pressure signals recorded by each device are elaborated to extract 10 feature variables which are used to segment the gait cycle into 6 sub-phases following a simplified version of Perry's gait model. The method is based on a Hidden Markov Model with a minimum phase length constraint and a univariate Gaussian emission model, which is decoded using a classic Viterbi algorithm. The method is tested on a pool of 5 healthy young subjects walking at two different speeds, through a leave-one-out cross-subject validation. The results show that the method is highly effective, yielding to an average performance of about 95% of correct phase classification, and 85 to 90% of phase transitions detected inside an acceptance window of 50ms.
De Rossi, S., Crea, S., Donati, M., Rebersek, P., Novak, D., Vitiello, N., et al. (2012). Gait segmentation using bipedal foot pressure patterns. In Proceedings of the IEEE RAS and EMBS International Conference on Biomedical Robotics and Biomechatronics (pp.361-366). IEEE [10.1109/BioRob.2012.6290278].
Gait segmentation using bipedal foot pressure patterns
Carrozza M. C.
2012
Abstract
We present an automated gait segmentation method based on the analysis of foot plantar pressure patterns elaborated from two wireless pressure-sensitive insoles. The 64 pressure signals recorded by each device are elaborated to extract 10 feature variables which are used to segment the gait cycle into 6 sub-phases following a simplified version of Perry's gait model. The method is based on a Hidden Markov Model with a minimum phase length constraint and a univariate Gaussian emission model, which is decoded using a classic Viterbi algorithm. The method is tested on a pool of 5 healthy young subjects walking at two different speeds, through a leave-one-out cross-subject validation. The results show that the method is highly effective, yielding to an average performance of about 95% of correct phase classification, and 85 to 90% of phase transitions detected inside an acceptance window of 50ms.
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
