Research On Sitting Information Recognition Based On Wireless Body Area Network

People spend most of its time sitting in the chair.There is sufficient medical data to show that maintaining a sitting position for a long time can lead to serious health problems.At present,people do not know much about their sitting position information,and can not intuitively know their sitting condition.In the current study,the extraction of sitting information is not enough,and there is no effective notification and intervention of poor sitting condition.To solve the above problem,this thesis designs a recognition system of sitting information.The main works are listed as following:(1)This thesis first searched many essay about the pathology of sitting posture,so as to understand the principle of harm caused by sitting posture.And a pressure cushion is designed to recognize the posture information of the human body,A wireless body area network(WLAN)system is also constructed,and the information will be transmitted to the intelligent terminal through wireless transmission.As for the cushion,thin film pressure sensor is selected as the core sensor,and we also improve the hardware devices based on ergonomic.(2)In order to enable the sedentary people to know their sitting condition in details,this thesis has designed an algorithm for recognizing sitting information,and distinguish the bad sitting postures and good sitting postures,and we also identify the specific sitting posture.The method used in this thesis is based on the time domain analysis.Firstly,the features of different sitting postures need to be extracted,then the extracted features are trained by variety of classifiers,and to test the validity of the extracted features.(3)This thesis also designs a method for recognizing the torso curvature using a pressure cushion.First of all,we analyzes the variation of the data of pressure when the torso is bending more and more,and reflects the bending condition of the human body by setting the corresponding characteristic value,then we maps the actual bending degree measured by the bending degree sensor,so as to realize the purpose to estimate the torso curvature based on the cushion.(4)This thesis also extracts the information about people's sitting postures.The main quantitative indicators include: the time of bad sitting postures,the time of good sitting postures,cumulative seating time,number of sitting posture changes,and number of sitting posture interruptions.And the corresponding algorithm is designed to extract these quantitative indicators.In order to avoid physical injuries caused by sitting behavior,this thesis also make appropriate intervention.By using the method of fuzzy logic to fuse the seating time with the time occupying ratio of bad sitting postures,then the integration of these two factors can accurately determine the intervention strategy.(5)Experimental analysis was also designed to verify the validity of the eigenvalues.Multiple classifiers were used to train the collected eigenvalue samples,And we use ten cross-validation method to verify the accuracy of the classification.At the same time,this thesis also analyzes the accuracy of the recognition of sitting position information.Experiment results show that the accuracy of recognition for sitting posture is 93.0%,which shows that the feature can well classify the sitting posture.And we also do experiment to verify the accuracy of recognize the torso curvature using a pressure cushion,and the comprehensive identification rate is 76.2%.And we also do experiment to verify the accuracy of the quantitative indicators,the comprehensive identification rate is 90.7%,so the system can effectively extract the information of sitting behavior.