Research On Application Of Kinect Motion-sensing Technology In Muscle Rehabilitation Training System

From about 30 years old,human muscle mass decreases at an average rate of 1-3% per year,which leads to muscle functional decay in the elderly,sequentially cause diabetes,fall and injury,muscle or joint pain,osteoporosis more likely,and brings difficulties to the daily life of the elderly.A weak leg muscle increases the risk of a fall.According to statistics,the sixth biggest cause of death of older people over 65 years old is caused by an accident caused by lower limb muscle weakness.Therefore,strengthening lower limb muscle exercise is particularly important for the elderly.With the aging population in China,the quality of the elderly’s life has been paid more attention.In the field of lower limb muscle rehabilitation,artificial assisted rehabilitation and machine-assisted rehabilitation are two common methods.These methods may require on-site guidance from professionals,or specific equipment and appropriate places,and high training costs,which bring many limitations to the muscle rehabilitation training of the elderly.In this paper,introduced human-computer interaction into the rehabilitation training,which brings great convenience to the muscle rehabilitation training of the elderly.For delaying the senile lower limb muscle decline training,adopted the theory of pattern recognition,Kinect bonetracking technology and other technical methods,through the Kinect platform,developed and completed the senile lower limb muscle decline training system.According to the needs of specific people,designed the training types and movements to developed a human-computer interactive training system for lower limb muscles rehabilitation.By using Kinect sensor hardware equipment and Kinect SDK bone-tracking API,collected human action and posture data,and calculated the angle and distance features of key joints according to the three-dimensional coordinate data of standard human lower limb posture and motion joints,and then analyzed the reliability,invariance and computational complexity of this two features.In this paper,use feature vector instead of three-dimensional coordinates of joints to describe human behavior,which solves the problem of low accuracy and differ form person.Then use the pattern recognition theory and method,build the static and dynamic standard training database,and realized the recognition of trainer’s posture and action by similarity matching algorithm based on NN and the threshold setting DTW algorithm under Visual C# development environment.Verified through the experiment that the recognition accuracy rate reaches more than 95%.Adopted the Microsoft WPF platform to design the user interface,formulated the training process of different training types.Ordered the evaluation indicators according to the pattern recognition method,and given the movement guidance and correction tips to trainers in time.At last,the test results of this system from the elderly show that it has a reliable recognition effect and a certain training effect.