• principal component analysis
• neuromorphic approach
• movement intention detection
• lower-limb assistive exoskeleton
• real-time algorithms
• robust classification methods

Population aging is nowadays one of the greatest challenges our society has to face. In the last years, wearable active orthoses have been developed with the purpose of providing elderly people with devices capable of assisting them in their motor activities. In order to deliver proper mechanical assistance, the user’s movement intention has to be identified. In this work, we propose two locomotion mode classification algorithms for the Active Pelvis Orthosis (APO) developed at the BioRobotics Institute of the Scuola Superiore Sant’Anna. The first algorithm (the “PCA-based” algorithm) makes use of the Principal Component Analysis (PCA) for pattern recognition. The other one, which we refer to as the “neuromorphic” algorithm, is based on a completely novel approach that converts the analog signals from onboard sensors into spikes and identifies the current locomotion mode by means of neuromorphic computational models. Both algorithms have proved to be effective and capable of achieving very high accuracy rates (well above 90 percent) with a short classification delay. The neuromorphic algorithm was then implemented in real time on LabVIEW®, integrated in the real-time controller of the APO and validated by means of an experimental protocol. It has proved to work properly in classifying the locomotion mode performed by the wearer. Furthermore, performance in a human-in-the-loop scenario, i.e. with assistance being delivered coherently with the detected modality, has shown high accuracy as well. Experimental evidence has shown that the neuromorphic algorithm could be effectively integrated in the control system of the APO for assisting ambulation in activities of daily living.