Advanced Analysis of 3D Kinect Data: Supervised Classification of Facial Nerve Function via Parallel Convolutional Neural Networks

In this paper, we designed a methodology to classify facial nerve function after head and neck surgery. It is important be able observe the rehabilitation process objectively specific brain surgery, when patients are often affected by face palsy. The dataset that used for classification problems in study only contains 236 measurements of 127 complex observations using most commonly House–Brackmann (HB) scale, which based on subjective opinion physician. Although there several traditional evaluation methods measuring paralysis, they still suffer from ignoring movement information. This plays an role analysis paralysis limits selection useful features paralysis. present triple-path convolutional neural network (TPCNN) evaluate problem mimetic muscle rehabilitation, observed Kinect stereovision camera. A system consisting three modules landmark measure computation parallel structure quantitatively assess considering region temporal variation sequences. proposed deep analyzes both global local patient’s face. These extracted high-level representations then fused final experimental results have verified better performance TPCNN compared state-of-the-art learning networks.