The Research And Application Of Information Fusion Based On Multi-agent

With the popular of artificial intelligence technology,agent technology has become a hot topic in the field of distributed artificial intelligence in recent years.The agents with autonomy and cooperation are capable of handling complex,collaborative and unpredictable problems,adapting their goals as the environment changes,learning and improving itself.However,the performance of a single agent is constrained by its resources,capabilities,and relationships with other agents,and can not solve complex problems alone.Therefore,the multi-agent system emerged.To introduce the information fusion technology into multi-agent systems,the local information perceived by the agents in space,time and function are fused with appropriate information fusion method,so as to obtain more complete judgments,assessments and decisions.To this end,this paper designs a kernel canonical correlation analysis information fusion model based on multi-agent,and has carried out the following work:(1)This paper analyzes the domestic and foreign research status and development trend of muti-agent system and information fusion,introduces the classification of agent model as well as the architecture of multi-agent syste.The multi-agent information fusion model based on the distribution of function,time and space is proposed.(2)Taking the audio recognition agent as the research object,this paper analyzes the extraction and combination methods of the voiceprint features used commonly,applies the deep learning to the voiceprint discrimination.A voiceprint discrimination model based on deep belief network(DBN)is designed.(3)Taking the video recognition agent as the research object,this paper introduces the common methods of video information recognition,especially the method of convolutional neural network,and the LSTMs are applied into the CNN model to improve the performance.A CNN video recognition model based on LSTMs is designed.(4)This paper analyzes the common methods of information fusion,introduces and improves the information fusion method based on CCA and the fusion method of KCCA is applied into the fusion model of nonlinear features.(5)The KCCA is introduced into multi-agent system.The KCCA information fusion model based on multi-agent design is applied into the speaker recognition and compared with the CCA information fusion model based on muti-agent,audio recognition agent model and video recognition agent model.The results of the experiments show that KCCA information fusion method based on multi-agent can identify the target effectively.The correct rate,anti-noise performance and stability are good.This method provides a way to achieve high-performance of target recognition in complex environment.