SVM based approach for the assessment of elbow flexion with smart textile sensor

This paper presents a new approach for monitoring the elbow flexion with a smart textile based on the piezo-resistive effect. The sensor is composed of a mix of conductive threads made of stainless steel fibers doped with metal particles and dielectric silk threads. The threads are arranged in a double bridle crochet structure to provide the best compromise between mechanical elasticity and sensor sensitivity. The sensor was integrated to a sweater to measure the joint angle of the elbow. In order to asses the angle in robust and accurate way, we developed a static model for angle recognition based on a machine learning algorithm: a Deep Neural Network. The design process of the static model includes the data collection, the extraction of features and the choice of the best architecture for the Deep Neural Network. This approach allow to reach recognition accuracy between 76.9% and 83.66%. Finally, a comparative study have also been performed in order to compare the performances of our model with different algorithms such as SVM, NB and DT.