Research On Intelligent Perception Technology Based On Vibration Signal

The author believes that in the upcoming smart world,sensors are everywhere,such as embedded in the ground and walls,and in the user's wearable device,thus being filled with thousands of invisible buttons.When the user changes his or her status or enters a specific location,the corresponding button is automatically triggered.From watches to credit cards to cars,roads and even entire cities,they can perceive human behavior and act accordingly.The author first built an indoor positioning system using vibration sensors on the ground,using pedestrian footstep vibration to locate pedestrians.The system first collects ground vibration signals through raspberry pies and geophones,then uses the Butterworth high-pass filter to filter out background noise,and then uses SWIM(Speed-based Adaptive Weight Increment Model)model to distinguish between footsteps and other vibrations noise,finally calculate the time difference and use TDOA(Time Difference of Arrival)to locate pedestrians.The system also detects ground vibrations,extracts Mel Cepstral coefficients as features,and uses Q clustering algorithm and random forest algorithm to identify pedestrian identity and behavioral activities,such as fall event for elderly.In addition,due to the small size of the sensors,the sensor interface for the user is lacking.The author proposes to localize the finger tap induced vibration on the surface of the body skin to provide an interactive interface for the sensors.In this paper,the piezoelectric ceramic piece and the accelerometer,gyroscope in the smart watch are used to collect the vibration signal.Then the system extracts the time domain and frequency domain features of the vibration signal.Based on the feature related point and the feature sensitive point,the features are optimized.At last,the system reuses the nearest neighbor algorithm and artificial neural network to identify different tap positions.Further,the system utilizes a density-based classifier to authenticate users.In this paper,the hardware system is built for the above two systems,and the effectiveness of the system is verified by real-world experimentalenvironment.The average localization error of the indoor positioning system is 7cm,and the recognition accuracy of the finger tapping position recognition system is 96%.At the same time,the user can be authenticated by a single tap at a very low equal error rate(2.4%),and the system shows strong robustness under various and real scenes(such as subways and airplanes).