Research On Signal Recognition Of Automobile Tailgate Based On Machine Learning

With the popularization of family cars and the rapid development of in-vehicle electronic equipment,while smart in-vehicle electronic equipment is recognized by people,car users have more advanced requirements for smart in-vehicle equipment.At present,as the key component of the smart car tailgate signal classification algorithm,mainly based on rule-based algorithms,lacking the response ability to user’s diversified operations.In this paper,the method of machine learning training model is applied to the recognition of automobile tail door signal..It is the first to try to use the OS-ELM algorithm to train the tailgate signal model online,and to apply the model in the research field of time series classification at this stage.Improvement and fusion,so as to obtain a new better tailgate signal classification model,and through experiments to prove the effectiveness and superiority of the tailgate signal identified model raised in this paper.The following part is the main content of this paper:(1)Due to the limited learning resources,there is no standard car tailgate signal data set.The car tailgate signal data set is collected by the author and related volunteers.The types defined as the start signal include: fast kick signal,slow kick signal,Standard kick,sweep kick,and bend knee kick,defined as interference signals including but not limited to: small animals walking,pedestrians approaching or standing away,foreign objects under the car,car bumpers contacting obstacles,etc.signal.The ratio of male to female signal collectors is 5:2.During the experiment,the clothes are diversified,including leather shoes,sports shoes,canvas shoes,etc.The data set will be used for the algorithm verification of this experiment.(2)Traditional machine learning has offline training and online classification work modes.After the model training is completed,the classification work is performed directly.In the offline stage,if new data is involved in training,the old data and new data need to be re-formed into a training set for model training.The online OS-ELM algorithm can perform classification work after the initialization model is completed,and the model can accept new data samples for model modification at any time.Its working principle is to pick up the period and frequency scopes characteristics of the capacitance signal.Input the vectors into the OS-ELM training model in batches.Experiments show that the accuracy of the model will increase as the batch of training samples increases.The OS-ELM algorithm used in this article is based on the classification of car tailgate signals.It is more suitable for car tailgate signal classification than classic machine learning algorithms.After multiple batch iterations of the classification model,the classification effect is better than the classic ELM model.(3)Use the Dynamic Time Warping(DTW)algorithm to compute the correlation between the capacitance signal participating in the training and the standard template signal set in advance,and dynamically tune-up the studying tempo of the recurrent neural network LSTM according to the different correlation,the purpose is to train The standard template signal is the main recognition target,and the other kick signals are the secondary recognition targets.In the same data set and the same experimental environment,through experiments to contrast the KNN model,K-means gathering model,and SVM model,it is found that the LSTM model combined with the dynamic time planning DTW algorithm is in the tailgate It performs better in signal experiments,and satisfies the purpose of the experiment.