A Study Of Position Mutation Based On Triaxial Accelerometer

In recent years, as the aging process aggravated, particularly "empty nesters" phenomenonas a result of one-child families which has increased in china, health problem of elder wasserious which caused by position mutation. As an example of falling, it has been a serious threatof old people to their health or even life. Health problem of elder due to falling caused a heavyburden to society every year, it has an important significance to identify falling accurately andalarm in time, which can rescue elder to reduce disability rate and mortality, body motionacceleration signal in database obtained by MMA7260QT is the object study, in introducing thebackground and significance of research status related fields.We got the body acceleration signal based on MMA7260Q, and stored into the database. Inintroducing the background and significance of research, the main contents are as follows:In the part of getting the body acceleration signal, we established the body space coordinatesand install the system. First, the body acceleration signal seems like a three-dimensional vectorof body space coordinates, we could use three axes to indicate the direction, so that X directionrepresents the vertical axis, Y-axis represents the longitudinal direction, Z-axis represents thelateral direction. Second, we wear the system in the heart of the left chest, and indicate thesampling frequency of the system on200HZ, the range of MMA7260Q select4g.In the part of pre-treating body acceleration signal, we get the body motion accelerationsignal and eliminate noise. The signal which was outputted by sensor was constituted by Earth’sgravity and body motion acceleration. We separate the body motion acceleration. Sensor outputcontained large impulse noise. We selected suitable method to eliminate it.We use vector SVM and vector SMA to describe the station of body motion to detect man’sfall, which has a good accuracy and real-time performance. Due to the higher position, the signalcontains amount of noise. Adopting the method of Bior3.3wavelet analysis, this thesiseffectively removes noise interference with characteristics and retained the maximum details ofoutline. When the body falls, SVM value exceeds0.9g, the process usually around60ms. Thesystem set the sampling frequency at200HZ, so if12consecutive sampling points are greaterthan or equal0.9g, the system repute that the body was fall. Fall in the human body detectionalgorithm proposed in this thesis greatly reduce the misjudgment ratio and false negative rate.In order to distinguish human daily activities (ADL) and fall, the first level fall detection isto judge whether the SVM is more than the detection threshold. On this basis, this thesis detectedwhether the SMA of the falls in the first level is more than threshold, to distinguish fall downand suspected. When the above two steps detection both judge that fall occurs, the system alerts. After a large number of simulation tests and analysis of test data, we can consider that SVMand SMA has a very low rate of false positives and false negatives.Finally, the detection method summary and outlook, although you can prove SVM andSMA algorithm can provide not only intuitive and effective quantitative indicators, however,little research in this area, further research work can be carried out more data added detectionmethod the amount of improvements and new features to explore, and so on.