Research On Performance Optimization Of Pressure Sensor And Sleeping Posture Recognition Method

The bedsore is a common clinical complication,which often occurs in bedridden patients.Effective recognition and monitoring of patients’ sleeping posture can help nurses adjust patients’ sleeping posture in time to reduce the risk of bedsore.However,the current methods have many problems,such as divulging privacy,binding human body and low recognition accuracy,which make it difficult to achieve clinical application.In order to obtain a higher recognition accuracy,the higher quality sleeping posture images are expected to be obtained by optimizing the performance of the sensor and data acquisition card based on the flexible sensor array,the preprocessing and feature extraction methods of sleeping posture images are focused to develope a better algorithm.This paper mainly includes the manufacture of the pressure sensor,the design of data acquisition card,the preprocessing and feature extraction methods of sleeping image and the development and test of the sleeping classification algorithm and the preliminary application of the sleeping monitoring system.The specific work are as follows:(1)Based on the flexible pressure sensor developed in this project,a large area flexible pressure mattress was made and a sleeping posture acquisition system was designed to collect the pressure image of human body.The main work in this section includes the design of the peripheral circuit of the acquisition system and the improvement of the signal-to-noise ratio by optimizing the core control circuit,power module,filter circuit and so on.(2)In this section,a targeted image preprocessing method is applied to obtain higher quality images,a feature extraction technology is applied to obtain a better data set for recognition.Many image enhancement including gray histogram analysis,image inversion and histogram equalization are applied in preprocessing process to remove the noise in the images.The training recognition data set is obtained by combining the extracted HOG feature and GLCM feature.(3)In the aspect of sleeping recognition algorithm,based on the principle of artificial neural network,a corresponding network structure is designed and the number of hidden nodes is optimized through a test.Using the optimization parameter,the sample set is tested with the network,the results shows that the highest recognition accuracy can achieve 99.17%.Compared with other sleeping posture recognition algorithms in the recognition method,the number of sleeping posture and the recognition accuracy rate,this method has some advantages.(4)In the initial application of the sleeping posture monitoring system,based on the study of the internal relationship between the intention of turning over and the execution of various movements of skeletal muscle groups,as well as the corresponding relationship between the movement of each skeletal muscle group and the dynamic pressure,the mechanical model of the back lifting and turning over(initially supine)of the human body are preliminarily established and the preliminary test is carried out.After the experiment in the prototype,the human body intention can be effectively judged,which lays a good foundation for the next roll over and other functions.