Research On The Non-contact Respiratory Detection Base On Flexible Pressure Sensor

With the aggravation of aging problem of our society year by year in China,the monitoring of the common respiratory disease in the elderly becomes more outstanding,and the modern medical research has proved that the major modern human disease is often associated with sleep apnea syndrome,and sleep monitoring is one of the most effective ways to obtain physiological parameters.But the effective method currently used is bound and uncomfortable to the human body,and the method operation is complicated and expensive,what is more is that it cannot be used for long-term monitoring and disease prevention.Seeking unbound detection methods has started at home and abroad,but most of them are in the relatively primary research stage.Therefore,this paper makes the unconstrained detecting respiratory method as the subject,and deeply studies current theoretical difficulties of sleep posture recognition based on pressure image by using fuzzy-rough set theory and accurate signal extraction in chest and abdomen,including feature representation and extraction of posture pressure image,sleep posture recognition based on pressure image by using fuzzy-rough set theory,the local breath signal extraction and so on.This paper's purpose is to solve the existing problems and be able to seeking the method to substitute detecting respiration method with bind belt sensor in the chest and abdomen.The main contents are as follows:1.Image Feature Representation and Extraction: The sleep posture pressure image can be obtained by using large flexible pressure sensor system,and the original pressure noise contained is eliminated in the image sequence through the study of the common image processing.Then,12 geometric features are selected to completely express the posture press images acted on the mattress lay and are explained in detail,according to the characteristics of posture pressure image.2.Sleep Posture Recognition: In the fuzzy rough set algorithm,a decision table that expressed the image knowledge information system is established for image collection,image features and image classification,the characteristic values for the recognition are analyzed and the continuous pressure distribution is discretized and converted to fuzzy decision table to identify sleep posture effectively by eliminating redundant information and inferring fuzzy decision rules.Experimental results show that the proposed algorithm can achieve a high accuracy.3.Local Breath Signal Extraction: Firstly,the human body parts recognition based on pictorial structure models is described in,which are used to dynamically capture the chest and abdomen positions.Then the respiratory signal is extracted from dynamic pressure data by using a vertical weighting algorithm in local position and eliminated the noise pollution to improve the quality of signal reduction through the noise processing;The discrete wavelet transform is applied to further eliminate the drift of the original breath signal,a more complete and clear respiratory signal is obtained by the smoothing processing,and the respiratory rate is extracted.