Gait Classification using LSTM Networks for Tagging System

Gait analysis based on observational methods is commonly used in everyday clinical practice. It is largely subjective and has a variety of limitations. In this publication we describe application of Recurrent Neural Networks (RNN) type Long Short-Term Memory (LSTM) for classification between healthy gait and gait with disease entity. This classification is based on person single step, beginning and ending with right foot strike on ground. The classification was carried out on reference motion database from Human Motion Laboratory (HML) of the Polish-Japanese Academy of Information Technology (PJAIT) containing a large set of samples. LSTM network checked in publication is a part of motion tagging system based on the EXTag concept.