An Efficient Dynamic Regulated Fuzzy Neural Network for Human Motion Retrieval and Analysis

Human motion retrieval and analysis is a useful means of activity recognition to 3D human bodies. An efficient method proposed estimate by using symmetric joint points limb features various parts based on regression task. We primarily obtain the coordinates located waist hip points. By introducing three critical feature torso points’ matching video sequences, its asymmetric will not be affected shading interference different postures. With parts, dynamic regulated Fuzzy neural network (DRFNN) for postures learning algorithm parameters weights. Finally, sequential actions corresponding are presented according best results DRFNN action database. Experiments show that compared with traditional adaptive self-organizing fuzzy (SOFNN) model, has higher estimation accuracy better presentation existing algorithms.