Research On Sleep Posture Recognition System Based On Computer Vision

Recognizing for sleep posture is an important part in the field of computer vision and pattern recognition. With the ageing of population, many rest home cannot be care elderly for twenty-four hours. Through real-time monitoring and identifying the old man sleep posture can capture for the motion of suddenly fell to the ground and make the corresponding treatment, such as call an ambulance or notify the guardian. Sleep posture has a great relationship with health, different people, especially the sick people, using appropriate sleep posture can prevent disease, so you can design a warning system based on sleep posture recognition. Therefore, sleep posture recognition have very important research and application values.There are few researches through the computer vision to recognition the sleep posture at home and aboard and nearly have not a complete research. Similar to other based on computer vision recognition, the method of sleep posture recognition contains two categories:sleep posture recognition in static picture and video. This paper contains following research and implementing on static picture sleep posture recognition:(1) Defined four basic sleep posture which is include supine, prone, left side and right side, this paper mainly in recognition of the four sleep postures.(2) Some pretreatment methods for sleep posture recognition are researched and implemented. In different pictures, there are various background, size and light conditions. So, histogram equalization and many other pretreatment methods are used to pretreated the pictures in this paper, which builds a good base for the sleep posture recognition.(3) Segment the target of the sleep posture pictures are researched and implemented, the target is the human of the picture. Because the image background and shape are complicated, traditional segmentation algorithm can not accurately judge the actual object boundaries. This paper use level set method segmentation of human body region in the image, compared with the traditional image segmentation method, level set method can segment the object with complicated shape in image since the evolving curve(surface) implicitly represented by the zero level set can naturally change its topological structure.(4) Feature extraction and recognition of the position after image segmentation is implemented. Because BP neural network has a good effect on target classification and recognition, so using BP neural network method designed the classifier and trained. The experimental results show that this method in the paper can adapt complex background sleep posture recognition, which not only have higher calculation speed, but also have good recognition effect.