Research On Behavior Recognition Method Based On 3D Convolutional Neural Network

In recent years,with the continuous development of computer technology,intelligent monitoring has been widely used in various areas of daily life.Its wide application scenarios and huge application value make behavior recognition a hot research topic in the field of computer vision,and the development of related technologies still has much room for improvement.Behavior recognition methods are mainly divided into two methods based on artificial design features and deep learning based methods.Due to the low efficiency and limitation of traditional artificial feature extraction behavior recognition,the deep learning-based convolutional neural network algorithm has attracted the attention of researchers with its characteristics of self-learning model features.Compared with traditional methods,the accuracy and efficiency of human behavior recognition based on deep learning have been greatly improved.Based on the 3D convolutional neural network,this paper conducts research based on the theory of pattern recognition and machine learning.First,the confidence rule base(BRB)algorithm is used to evaluate the hyperparameters.The hyperparameters here refer to the parameters set before starting the learning process.Not the parameter data obtained through training.Such as the learning rate,the number of batch samples,the number of convolution kernels,and so on.Then,the particle swarm optimization algorithm(PSO)is used to optimize the hyperparameters of the 3D convolutional neural network.The aim is to improve the 3D convolutional neural network,save the time of deep learning model training,improve the effect of behavior recognition,and improve the robustness of the algorithm.The main research contents and contributions of this article are as follows:1)Based on the 3D convolutional neural network,for the long training time of deep learning,a BRB-based hyperparameter evaluation method is proposed.The BRB model uses qualitative knowledge to establish initial belief rules,and uses quantitative knowledge data to optimize the initial parameters of the BRB model,so that this parameter can effectively use various types of uncertainty information.The experimental results show that the BRB model is very close to the real value,that is,this model is fitted with the 3D convolutional neural network model,and at the same time,a high evaluation accuracy is obtained,which verifies the effectiveness of this model.2)In order to find the optimal initial parameter value,this paper uses the particle swarm optimization algorithm to optimize the hyperparameters of the 3D convolutional neural network.In the optimization process,hyperparameters are randomly generated,and then the hyperparameters are used as the input of the BRB model,and the evaluation model is used as the fitness function to output the ideal hyperparameters.This method saves a lot of time compared to taking the 3D convolution model as the target optimization function.Finally,the obtained parameter values are used to verify the 3D convolution model.The experimental results show that the method improves the efficiency of 3D convolution model behavior recognition and classification effect.