A Research On Auto-adjusting Mattress System Based On Image Technology

In medicine, thoracic kyphosis and the deformation of spine is known as kyphosis. As we all know, sleep is essential for human health, but for hunchbacked man, it is difficult to sleep in supine position, resulting in a lot of pain of the hunchbacked man. Especially for the patients of severe kyphosis, when lying on the common mattress, resulting in extrusion of the protruding parts, even producing pressure ulcers, bone injury.Therefore, this paper specially designs a mattress system which can automatically adjust itself for the hunchbacked people based on image technology.According to the sleeping postures of patients, this paper designs an auto-adjusting mattress to control the inflating and deflating process of balloon, which is located under the points of the patients’ hump salient. According to personal features of the patients’ physiology and pathology, we designed the mattress. First, transform an ordinary mattress, and dug a rectangular hole size of 45 * 42cm2 in the middle of the mattress, in which place the die pad sponge which matches and the patient’s spine, according to the position of each patient’s hump salient points. Then using use special cutting machine, the sponge pad is made according to the distribution value of the patients’ supine pressure, left lateral pressure, right lateral pressure through biomechanical modeling and CAM processing, and place the balloon in its bottom, connecting the two pump through the microcontroller Samsung S3C6410. Finally, the system obtain an image of the patient’s sleeping position from the near-infrared camera every 1 minute, and make the detected image to be geometric normalized, histogram equalizated, image filtering and Haar-like method extracted. and send the recognition results to the embedded systems, controlling air bag to be inflated or deflated as well as supporting the patient to change posture.The system uses BP neural network algorithm to determine the patient’s sleeping position, based on the recognition result, adjust the airbag, supporting a patient posture change. First, the system uses ORL face database and face samples in Poser software, and extract the rectangle features of front face and side face from these samples, then the feature information is sent to BP neural network for training, and change the weights until the output of the results match the expected results so as to build the BP neural netw ork classifier. Secondly, the system obtain an image of the patient’s sleeping position from the near-infrared camera and make the detected image to be geometric normalized, histogram equalizated, image filtering and use class Haar-like method to extract the pixel integral value matrix characteristics, and these characteristics of matrices are passed into trained BP neural network. Finally, If the output is 00, representativing the detecting result is frontal face, considering to be supine, and at the same time, adjusting the balloon to the deflating state. If the output is 01 or 10, respectively representing the detected face to the left and right sides of the face, thus the controlling balloon is inflated state, supporting patient on one side. If the output is 11, representativing that the above three cases is not detected and the system is without any treatment. The results showed that the system is of high recognition efficiency. Adjust the posture airbags according to the posture of patient at each real-time, thus supporting a patient po sture change.