Gait Phase Classification of Lower Limb Exoskeleton Based on a Compound Network Model

The classification of lower limb gait phase is very important for the control exoskeleton robots. In order to enable determine and provide appropriate assistance wearer, we propose a compound network based on CNN-BiLSTM. method uses data from inertial measurement units placed leg pressure sensor arrays sole as inputs model. convolutional neural (CNN) used obtain local key features data, then bidirectional long short-term memory (BiLSTM) extract serialized information high-level feature expression. Finally, seven phases both feet were obtained through softmax layer. We designed acquisition system collected subjects at varying walking speeds. test set, highest accuracy can reach 95.09%. compared proposed model with (LSTM) gated recurrent unit (GRU) network. experimental results show that average CNN-BiLSTM 0.417% higher than LSTM 0.596% GRU Therefore, ability classify be applied in designing controllers better assist different correctly wearer walk.