Introduction: Disturbances in balance control lead to movement impairment and severe discomfort, dizziness, and vertigo. They can also lead to serious accidents, due to the loss of balance in critical conditions. It is important to monitor the level of balance in order to determine the risk of a fall and to evaluate progress during treatment. Some solutions exist, such as those based on cameras and force platforms, but they are generally restricted to indoor environments. We propose and evaluate a system, based on accelerometers and support vector machines (SVMs), that indicates the user's postural balance variation by monitoring signals related to balance, and which can be used in indoor and outdoor environments.
Methodology: The proposed system consists of a second-skin shirt, six accelerometers, a 328 ATMEGA microcontroller, and a local storage module. For the training phase, we used the accelerometer signals acquired from a single subject under monitored conditions of balance and intentional imbalance, and used the scores provided by a validated commercial solution (the SWAY® software) for establishing the reference target values. Based on these targets, we trained an SVM to classify the signal into n levels of balance (with n varying from 2 to 7) and later evaluated the performance using cross validation by random resampling. We also developed an SVM approach for estimating the center of pressure based on the signals from the accelerometers, by using as reference targets the results from a force platform by AMTI®. We considered ve possible regions for the center of mass, and our system was used to determine the correct region using the accelerometer signals. For validation, we performed experiments with a subject who was rst standing, and later walking, performing a body rotation, and performing sudden intentional drops. Later the subject was requested to stand and then incline in four main directions, so the di erent centers of pressure (COPs) could be computed by our system and compared to the results from the force platform. We also performed tests with a dummy and a John Doe doll, in order to observe the system's behavior in the presence of a sudden drop or a lack of balance.
Results and Discussion: The results show that the system can classify the acquired signals into two to seven levels of balance, with success rates ranging from 92.5% (for seven levels) to 98.3%, in 1000 sessions of random resampling. With two levels of balance, the system attains in the best case an accuracy of 98.9%. The average accuracy with two levels of balance was signi cantly greater than 93% ( p =0.045) and the accuracy was signi cantly greater than 97% ( p =0.044). With seven levels of balance, the accuracy was signi cantly greater than 94% ( p =0.046) and the precision was signi cantly greater than 80% ( p =0.049). The tests performed with the dolls show that the system is able to distinguish between the conditions of a sudden drop and of a recovery of balance after losing one's balance. In this case, the average accuracy was greater than 95% ( p =0.043) and the precision was greater than 95% (p=0.026). The system was also able to infer the centroid of each COP region with an error lower than 0.9cm (p=0.0045). These results suggest that the system can be used to detect variations in balance and, therefore, to indicate the risk of a fall even in outdoor environments.