Research On Technology Of Travel Pattern Recognition Based On Deep Learning

In recent years,with the improvement of mobile Internet and embedded technology,human behavior recognition based on smart mobile devices has become a popular area of ubiquitous computing.As one of the most important research directions,travel pattern recognition can not only provide essential data for urban planning and traffic management,but also have positive influence on the development of scene awareness and intelligent environment.At the present stage,travel pattern recognition methods based on mobile devices mainly adopt GPS,accelerometer,Wi-Fi and other data,using traditional machine learning algorithms such as AdaBoost and SVM to classify and recognize travel patterns.Generally,many problems reflected in present researches,such as large energy consumption,unsatisfactory precision and reliance on complicated hand-crafted features,and it is still a big challenge to make a fine-grained detection of motorized modalities.The application of deep learning in the field of pattern recognition is a hot topic in recent years.The basic idea is to establish a multilayer neural network simulating human brain and approximate nonlinear complex function,which has shown strong characteristic learning ability in the research of user behavior detection and traffic pattern recognition.Referring to current researches on travel pattern recognition at home and abroad,this paper presents a travel pattern recognition method based on deep learning.This method obtains users' transportation travel data using sensors with low-power consumption,including accelerometer,gyroscope,geomagnetic sensor and barometer,and sliding window filtering and feature extraction are performed before sensor data is fed into the Multilayer Perceptron(MLP),Convolutional Neural Network(CNN),Long Short-Term Memory(LSTM)to learn deep features of motorized vehicles.During the experiment,various important hyper-parameters such as number of hidden layers,convolution kernel size,number of neurons in each layer,activation function and so on,are monitored and modified to adapt the characteristics of multiple sensor signals.In addition,optimization algorithms such as dropout,momentum and adaptive learning rate are used to speed up the convergence of the model.In order to improve the stability and robustness of the system,the Bayesian voting algorithm is used to fuse the results of different neural networks.In this paper,the algorithm is simulated and tested several times,and the experimental results show that the proposed algorithm can achieve 98%accuracy to distinguish between four kinds of motorized travel modes,including car,bus,metro and train,which outperforms other recognition methods based on machine learning.This paper provides a new method for the research of transportation travel mode,which has the value of further research and development in practical application.