| Many obstacles have happened in current in-vehicle voice user interaction(VUI)systems,such as failure to wake up,wrong execution intention,which occupy cognitive and attention resources for drivers,increasing accident risk.Therefore,it is essential to improve the collaboration between drivers and in-vehicle VUIs for safe driving and efficient driving.However,most of the existing researches on drivers’reactions to obstacles in in-vehicle voice interaction systems are subjective studies.Due to the lack of objective data,it cannot provide a clear direction for the optimization of voice interaction technology and it cannot provide clear guidance for the design of voice interaction system.This research aims to quantitatively evaluate the impact of different types of voice interaction obstacles on drivers through eye-tracking technology,find patterns and establish a model to help designers and evaluators design better in-vehicle voice interaction systems.This paper mainly includes the following three aspects:First,basic theoretical research:(1)Research on the types of voice interaction obstacles:Through literature analysis and product analysis,the types and principles of in-vehicle voice interaction obstacles studied in this paper are clarified.(2)Eye movement indicators research:Summarize the eye movement indicators and their meanings,eye movement data analysis methods.Establish an indicator set that can measure the driver’s state.Second,experimental research:(1)basic experimental research:Based on the Wizard-of-OZ platform,an eye-tracking device is used to measure the difference in cognitive load,mental load,and attention interference of drivers when they encounter different voice obstacles.Six eye movement indicators are found that can effectively measure this difference.It proves the effectiveness of eye movement technology in evaluating in-vehicle voice interaction systems and the different impacts of different voice interaction obstacles on drivers.(2)Eye movement data mining:Integrate various eye movement indicators to output the severity of these obstacles.Discuss the differences between subjective scale data and objective eye movement data in experimental results.Build the cognitive curve of in-vehicle voice interaction and the obstacle judgment model based on eye movement data.The model can predict four types of voice obstacles,and the accuracy is about 93%.Third,the application of the research results:(1)Use the in-vehicle voice interaction cognitive curve and obstacle judgment model established in this study to evaluate Siri and Xiaodu voice assistants and put forward optimization suggestions.(2)Combined with the obstacle judgment model,design an adaptive in-vehicle voice interaction system that can repair voice obstacles,including system structure and repair strategies.From basic theoretical research,experimental research to the application of the research results,this paper focuses on the impact of invehicle voice interaction obstacles on drivers by eye-tracking technology.And the pattern and model are explored.That provides an objective and effective methodology for evaluators and technicians of in-vehicle voice interaction systems from the visual and data perspective.The design suggestions and strategies of adaptive in-vehicle voice interaction systems are discussed to improve the human-computer collaboration ability in invehicle voice interaction systems. |
