| With the popularity of wearable equipment and the continuous development of pervasive computing,more and more research will focus on the acceleration sensor data up.Users and the terminal portable equipment gradually showing a tight coupling situation,at the same time,the smart phone computing power in recent years,a leap in the improvement,how to use a good mobile intelligent equipment,a high degree of carrying time and computing power is an important breakthrough.The acceleration sensor data as an important part of describing the information of human activities,the gait features,behavior patterns and other information carried by the human body for the understanding of semantic is of vital importance.User activity data collected through smartphones is not only time-long and large-scale,challenging data processing and storage.Therefore,this paper takes the end point detection of human activities as the starting point,and takes the starting point and the ending point of the human body activity acceleration data as the target,and proposes the endpoint detection algorithm under two different scenarios.Specifically,the work is as follows: Aiming at the limitation of the computing power and memory resources of the smart phone as a signal acquisition device,an improved dual-threshold human activity endpoint detection algorithm is proposed.And three different short-time zero-crossing rates are defined for the three-dimensional space vector data: the component axis is short-term direct zero-crossing rate;the spatial reverse zero-crossing rate;the zero-crossing rate based on the diagonal hexagrams.By comparing the computational complexity and the detection validity of the three,we choose the spatial reverse zero-crossing rate as the discriminant parameter.The algorithm can perform coarse-grained behavior detection,avoid uploading all data,save a lot of network transmission bandwidth and server-side storage resources.First,aiming at the limitation of the computing power and memory resources of the smart phone as a signal acquisition device,an improved dual-threshold human activity endpoint detection algorithm is proposed.Three different short-term zero-crossing rates are defined for three-dimensional space vector data.The algorithm can perform coarse-grained behavior detection,avoid uploading all data,save a lot of network transmission bandwidth and server-side storage resources.Second,in order to improve the detection strategy of human activity data(acceleration)under the condition of limited resources,the improved double threshold detection method of human activity endpoint detection algorithm is proposed.The strategy includes the corresponding dynamic sampling strategy,upload window decision,data storage queue and the establishment of the upload queue and so on.Through the transmission strategy proposed in this paper,can effectively reduce the transmission costs and data storage costs.Thirdly,aiming at the need of more accurate segment extraction in the process of specific human behavior recognition,a method of detecting the human body activity based on the entropy of the acceleration data is proposed.And in order to avoid the loss of the acceleration vector direction information when calculating the root mean square of the three-axis data,a three-dimensional acceleration source joint information entropy model is constructed.Compared with the double threshold algorithm,the algorithm is more accurate than the double threshold algorithm,but the detection result is more accurate,and it is suitable for extracting the actual activity data as the data preprocessing step in the early stage of human behavior identification.Through the verification experiment,it is proved that the proposed two-threshold method can effectively reduce the amount of data generation.The transmission strategy can save the transmission cost.The information entropy detection algorithm can effectively improve the accuracy of behavior recognition in complex cases. |
PDF Full Text Request