Research On Vibration Communication Of Smart Devices Based On Convolutional Neural Network

In recent years,smart Io T devices including wearable devices,smart homes and healthcare devices are booming,which has greatly changed people's daily life.People often need to allow different smart devices to exchange information.Commonly used methods include Wi-Fi,Bluetooth,Zigbee,etc.But they are all vulnerable to wireless eavesdropping attacks.Vibration communication based on smart devices uses a vibration motor embedded in a smart device as a signal transmitter and an accelerometer as a signal receiver,allowing vibration signals to propagate through a rigid solid surface,which solves the security problems of the abovementioned wireless communication methods.However,the accuracy,transmission rate and applicability of various current vibration communication algorithms on commercial equipment have certain limitations.This paper proposes a new vibration communication algorithm based on convolutional neural network,which greatly improves the transmission rate of vibration communication while ensuring the transmission accuracy.Unlike the existing vibration-based communication algorithm that uses threshold division and decoding,this algorithm uses machine learning to decode the vibration signal of a multi-ary smart device,where the message sender sends a symbol group composed of multiple bit groups Signal,and then use machine learning at the receiving end to decode the symbol group signal.This paper chooses convolutional neural network as the core engine of the decoding algorithm to learn and recognize vibration patterns.By collecting a large number of data sets for in-depth evaluation,the model based on Convolutional Neural Network(CNN)achieves the highest decoding accuracy of 97% when the algorithm transmission rate is 40bit/s.Under the same transmission rate,the traditional vibration communication algorithm only has a transmission rate of 10bit/s to 20bit/s.In addition,by evaluating the performance of the algorithm in this paper in a variety of challenging experimental environments,such as different placement directions of smart devices,different distances,and different types of devices,it shows that when CNN is used as the core engine of the decoding algorithm,The algorithm in this paper has better robustness.Finally,this paper also implements the system algorithm on existing smart phones and smart watches.By evaluating the time and power consumption of the algorithm implementation,it proves that the algorithm can be used in low-cost smart devices(such as smart watches).It runs in real-time without causing excessive resource consumption.