Human Posture Recognition Base On Accelerometer

With the rapid development of Micro Electro Mechanical System technology and the in-depth research of pattern recognition theory,human posture recognition based on accelerometer has gradually become an important research direction in the field of pattern recognition.It has received a widespread attention in the area of motion analysis,medical ward,somatosensory game and energy consumption.Compared with the human posture recognition based on image analysis,the approach based on accelerometer is not subject to environmental restraint,and the less cost and energy consumption make it has a wider application prospect.The current research on human posture recognition is still in the comparision basic stage.Due to the diversity of objective environment and the complexity of posture,this research still has a lot of urgent problems to resolve,such as how to extract data features with stronger representation abilities,how to design a classifier which has high accuracy and high efficiency.In order to solve the above difficulties,this paper focues on the human posture recognition based on accelerometer has launched a series of studies,the main tasks are listed as follows:1.The existing human posture recognition methods are summarized.The approach based on image analysis and that on accelerometer are compared.Systematic analysis the data acquisition module,data pretreatment module,feature extraction and selection module,classification algorithm module.2.An improved particle swarm optimization algorithm(PSO)for training the neural network is proposed to recognize the human posture.Based on the common feature sets,two feasures including the discrete coefficient and curve integral which can reflect the trend of acceleration and the variation of velocity are proposed,and these features are extracted as the input of the neural network.Controlled by probability,the particles are operated by genetic operator when the neural network is optimized by PSO algorithm and it will make the particles jump out of the local mininum value.Finally,the six activities are recognized through the trained neural network.The experimental results show that the optimized neural network based on modified particle swarm optimization algorithm can improve the convergence speed and the ability of global optimization.Compared with other classical algorithm,the proposed algorithm has higher recognition accuracy.3.A new algorithm based on the window similarity is proposed.This algorithm has fitting the discrete data at first,and obtained the fitting curve by taken the use of particle swarm optimization algorithm to optimize the key parameters.Then calculated the windowsimilarity between the different fitting curves.And we use the window similarity as the distance measure.At last we adopt the K-Nearest Neighbour classifier to identify the human body posture.The experimental results show that the proposed algorithm is effective to recognitize the 6 activities,and it has less computation cost.In conclusion,the current research on human posture recognition is still in the development stage.The research of this subject has important theoretical value and application requirements.So,it is worth carrying out more detailed and further research.