Towards Object Structure Detection By Multil Sensor Fusion On Smart Phones

With the continuous upgrading of smart phone hardware,the integrated CPU on the mobile phone makes the computing power and processing speed of the smart phone faster and faster,ordinary portable intelligent devices also began to equip with a variety of sensors,with more and more powerful environmental awareness.More and more applications developed by using various sensors embedded in smart phones appear in people's vision.Some common applications include indoor navigation,engine fault detection,brain wave detection,heartbeat monitoring,early detection of Parkinson's patients by detecting hand shaking,etc.According to the observation in our daily life,it is possible to detect the internal structure of the object by using the sampling ability of the inertial sensor of smart phone to the environment vibration echo.After a series of experiments and preliminary analysis,an object structure detection scheme based on smart phone is proposed to detect the internal structure of daily objects conveniently and quickly,detect the structure of some objects without damaging the internal structure of the object,and meet the real needs of people.The main work and contribution of this paper are as follows:First of all,the propagation characteristics of the vibration signal of mobile phone are studied.According to the propagation characteristics of the vibration signal,a two-stage interference cancellation scheme is proposed to effectively obtain the echo signal carrying the hidden object structure information in the vibration signal,and then the extracted echo signal is processed by moving average filter to remove the impact of random noise,which is the basis of the next signal analysis.On the basis of obtaining effective echo signal,the signal with characteristic value returned by the object is analyzed in time domain and frequency domain,and relevant features are extracted.After extracting and analyzing many features in time and frequency domain,the amplitude and power spectrum of signal in time domain and frequency domain are selected as the main features of acceleration sensor for object detection,and the average value of magnetic field intensity measured by magnet sensor is taken as the main features of magnet sensor for object detection.Then,a fusion method based on the data of accelerometer and magnet sensor is proposed to improve the accuracy of sampling and decision.Because magnet sensor is easy to be affected by many interference factors,in order to improve the accuracy of magnet sensor sensing objects,the least square ellipsoid fitting is used to calibrate the magnet sensor to remove the influence of hard iron interference and soft iron interference.In order to improve the speed of object detection,this paper uses distributed statistical decision fusion to realize the fusion of accelerometer and magnet sensor.Through experimental comparison and analysis,compared with using accelerometer and magnet sensor to detect objects,the fusion algorithm of accelerometer and magnet sensor based on distributed statistics has achieved better detection accuracy.Finally,on the premise of collecting a large number of sample data sets,SVM algorithm with better classification effect and faster processing speed is used to detect whether the object is metal and whether there is a cavity inside.The SVM classifier is trained in the training set,the SVM classification algorithm is used to classify and recognize the test set,the detection of the internal structure of the object is realized,the accuracy of the internal structure recognition of the object is obtained,and the results are compared and analyzed with the accuracy of the fusion algorithm of acceleration sensor and magnet sensor,the results show that the two algorithms have achieved a better(an ideal)detection effect.