Research On Auditory Attention In Human-machine Interactive System Concurrent Multi-speech Environment

Radical change has taken place in the field of human-machine interaction with the development of digital,automation and information technology.The operator's task in large centralized control centers,such as air traffic control center,main control room of nuclear power plant and rail transit control center,has changed from manual control to information supervision and control-oriented decision-making,the performance more dependent on attention,communication and other non-technical skills.Maintaining a high-level situational awareness on digital interface information is critical for human error avoidance.Dispatcher obtain more than 30% of the information through speech interaction tasks in rail transit control center.The successful acoustic design of control center could be controlled background noise within 45 d B(A)and eliminated obvious interference to operators.However,parallel tasks cause multi-speech concurrency,which obstructed the information transmission between dispatchers.Concurrent multi-speech will compete for limited attention resources,resulting in auditory response delays,information identification errors and even causing safety accidents.The topic of this dissertation comes from the independent subject of the National Key Laboratory of Rail Transit Control and Security: "Research on the Complexity of Dispatching Behavior in Rail Transit Control Center".The theoretical and practical value of this dissertation is to explore the bottom-up and top-down auditory selective attention behavior by psychoacoustic experiments based on the factors affecting auditory attention.The prediction model of auditory selective attention based on multiple criteria decision-making theory and the intervention measures of auditory attention based on auditory tracking task were studied,which provide theoretical methods and technical support for effectively solving the reliability of auditory information processing in concurrent multi-speech environment of human-machine interaction system.The main innovative work of this dissertation is as follows:1)Based on the Critical Incidents Technique(CIT)method,the factors affecting auditory attention were quantitatively analyzed through the design and implementation of the questionnaire.Thirteen factors affecting auditory attention were identified in three dimensions: auditory,cognitive and information.2)Based on the analysis of the information masking mechanism of perceptual cues,two hypotheses were proposed namely centralized exposure of speaker features and information intelligibility may improving target discrimination rate in concurrent multi-speech environments,and behavioral experiments were carried out to verify the hypothesis.3)On the basis of the interaction analysis of bottom-up and top-down factors on auditory selective attention,a model of auditory selective attention was established based on the multiple non-linear regression equation,in which target saliency and cognitive processing were combined.It was found that the probability of auditory selective attention was significantly correlated with the saliency and the information value of target.4)Based on the analysis of auditory training method in concurrent multi-speech environment were discussed,ergonomic experiments show that short-term centralized compound auditory tracking task based on persistent rhythmic stimulus could improve auditory attention,which provides a scientific basis for attention plasticity and transfer effect of attention in speech-in-noise scene.