• algorithm
• balance
• center of pressure
• gait analysis

Falls and loss of balance are a serious problem facing elderly people, amputees and subjects with neurological pathologies. Today the increasing life expectancy and medicine and rehabilitation progress let people with serious diseases walking in an autonomous way, but loss of balance or falls are a common risk. That is why the research is going to the direction to give the right support to prevent falls, searching strategies for pre-fall detection. This work wants to introduce a novel algorithm to detect loss of balance that could result in a fall, having only information from feet, provided by wearable foot insoles with pressure sensors, developed at “Istituto di BioRobotica, Scuola Superiore Sant’anna” in Pisa. Experiments were made in the lab of Gait Analysis at the TuDelft (Nederland) by a group of four researchers. The global experimental setup included IMU (Inertial Measurement Unit) on the upper body and sensorized insoles at the feet. Everything was recorded by motion capture’s system. After data collection I used only centre of pressure data from insoles to extract some features, after that I decided to use the SVM (Support Vector Machine) approach to train my data and classify movements in order to detect loss of balance during walking. This work grounds on the hypothesis that differences in features values extracted from CoP are found between normal walking and perturbed walking. Indeed during experiments it was asked ten subjects to perform normal walking on a treadmill at three differents velocities and walking while perturbations produced a loss of balance.