Research Of Wheelchair Users' Fatigue Detection System Based On Multi-sensors

Fatigue is one of the causes of the decline in normal working ability,many diseases and major accidents.Therefore,in recent years,more and more experts and scholars in different fields have focused on fatigue detection and prediction.In fact,the phenomenon of fatigue is ubiquitous throughout the society,but a majority of the experts and scholars takes the athletes or drivers as the research subjects.Many issues,that include the complexity of fatigue,the variant evaluation criterion,the simplicity of fatigue feature,and lack of multi-information feature fusion,lead to the generally low accuracy and poorly practical value of current fatigue detection.In order to solve the above problems,this paper builds a distributed Wireless Body Area Network combined with ECG,sEMG and acceleration sensors,and designs the fatigue detection and classification system integrated with fuzzy neural network.Then,the system has been validated by wheelchair users.It finally achieved a fatigue detection classification system with higher fatigue classification accuracy and can be applied to the actual life of the crowd as fatigue indication,thus helping this population to avoid the secondary damage caused by the decline in physiological sensory function.The main work and achievements are summarized as follows:1)The fatigue detection method has been designed.The objective and subjective fatigue evaluation method are combined to evaluate and study the fatigue of the human body in daily life.Physiological(ECG and sEMG)and kinetic signals(acceleration)are combined as the main fatigue evaluation method.Self-fatigue evaluation data is taken as an auxiliary evaluation.Thus,it achieves a multi-signal source fatigue determination which avoids the low accuracy caused by a single signal source.2)The fatigue classification system has been systematically constructed.The whole structure and working flow of the fatigue detection system are introduced.The hardware is described from three different layers-the acquisition layer,mobile end layer and data analysis layer.The layout of the sensor and the construction of the body network are completed.The software part includes that LABVIEW is combined with Shimmer plug-in as the secondary development of fatigue detection system.It achieves a smaller packet loss rate and a simultaneous collection and storage of multiple data,containing the ECG,sEMG and acceleration sensors and other sensors' data.Besides,it also achieves a fatigue display through the Small Love Assistant APP.3)The algorithm of fatigue classification system based on fuzzy neural network has been designed.Firstly,the framework of the whole fatigue classification system is introduced,including preprocessing,feature extraction,classifier training,as well as output of final fatigue results.The concrete algorithm is designed to remove the noise by using the Butterworth filter and the Wavelet Transform,etc.;to extract the effective EMG signal ECG signal ARV features combined with fragment time window and to extract features of MPF and LF;to realize the fatigue classifier based on the NEFCLASS;to design the input and output membership function,as well as fuzzy rules.4)A couple of experiments have been designed.And a number of key issues and the fuzzy neural system have been verified.Designed experiments include the data acquisition experiment and the muscle selection experiment.Experimental verification includes verification of the relationship between features and fatigue;the method of sEMG signal segmentation and reconnection;and the accuracy of trained fuzzy neural classifier.The fatigue detection system achieves 80% classification accuracy.Compared with the fatigue classification accuracy of the existing literature,the method and system proposed in this paper proves more superior.