Abstract
Motion intent recognition as a key technology to help wearers control the exoskeleton of lower extremities has received extensive attention in recent years. The surface electromyography (sEMG) signal of the lower limbs is the most commonly used identification signal source. The traditional recognition method is to extract the feature manually and then use the machine learning method to train the model. The recognition accuracy depends on the prior knowledge, and the manual extraction feature is more troublesome. This paper uses a CNN-RF combined model to recognize five movements (stand, sit, walk, up and down stairs). Convolutional neural network (CNN) has autonomous learning ability automatically extracts features and can combines with traditional random forest (RF) model for training. Firstly, real-time experimental data was extracted by four-channel sEMG sensor and gyroscope, then the convolutional neural network automatically extracted features, finally the feature vector was fed to the random forest model for training. The experiment achieved a high accuracy on the test set and the training speed also meets real-time requirements, which proved the superiority of the method.