基于認知特性的飛行員視覺注意力分配建模與實驗研究

莊達民 完顏笑如

摘 要：該研究以飛機駕駛艙人機界面設計為應用背景，通過結合具體飛行任務開展復雜人機界面環境下，飛行員視覺注意機制的數學建模研究，以期對飛行員的視覺注意力分配行為進行事前預測，先后提出了基于混合熵最大化的注意力分配模型和多因素條件下的注意力分配模型。并且根據工效學仿真實驗對所提出的注意力分配模型進行可用性驗證，從而為飛機駕駛艙的人機界面設計提供理論支持和技術手段。研究針對飛行員視覺注意力分配機制進行數學建模，并基于飛機仿真顯示界面開展工效學驗證實驗，結合紅外眼動裝置對注視點分布的測量，應用回歸分析的統計方法驗證模型。基于混合熵最大化的注意力分配模型是以飛行員的儀表掃視行為為主要應用情境而建立，以一組視覺加工信息為研究對象，以認知工程、模糊數學和廣義信息論為主要理論基礎。為驗證數學模型的可用性，根據注意力分配理論模型可計算出作業人員對各儀表信息的注意力分配比例理論值；通過開展注意力分配仿真實驗，注意力分配比例的行為學實驗值通過作業人員對不同儀表信息的正確操作率進行統計獲得；注意力分配比例的生理學實驗值通過作業人員對不同儀表信息的注視點數量進行統計獲得。通過將注意力分配比例的理論值與各實驗值進行對比，得出在腦力負荷適中情況下所建模型具有一定可用性的結論。雖然基于混合熵的注意力分配模型較好的量化了認知過程中的模糊性和隨機性，但是由于模型本身對于影響因素的結算需要對信息檢出效率進行事先測量，從而降低了模型的預測能力。為了進一步完善和優化注意力分配模型，在混合熵模型和SEEV模型的基礎上提出了多因素條件下注意力分配數學模型。多因素模型從基于飛行員的儀表掃視行為所反映的認知特性分析出發，密切結合航空任務情境，對前人的模型進行了重新的解釋、整合和詳細地公式推導，并結合之前在視覺信息編碼方面所開展的工作，將“信息檢出效率”這一因素也引入到注意力分配模型中，從而提出了一個新的注意力分配模型，該模型在混合熵的理論基礎上，綜合考慮了信息價值、人為失誤以及信息檢出效率等因素對飛行員注意力分配策略的影響。為將所建模型應用于航空信息顯示界面優化設計并驗證其有效性，基于飛機平視顯示器仿真模型，通過測量被試者對所監視視覺目標的識別績效，同時結合眼動儀的實時觀測數據，分析了被試者的實際注意力分配情況。實驗結果初步驗證了該注意力分配模型在復雜人機界面下的有效性。

關鍵詞：視覺注意力分配 數學建模 工效實驗 認知特性

Abstract：In this research， pilot attention allocation mechanism under complex human-machine interface environment was studied and analyzed in a mathematical fashion in the application background of aircraft cockpit interface design.To predict pilot attention allocation， a serial of mathematical models were established which contained one model based on hybrid entropy and the other under multiple factors condition. In validation of these models，ergonomics simulation experiments were designed and carried out to verify the effectiveness of the models，which provided both theoretical basis and technological method to optimal design of human-machine interface in aircraft cockpit.Quantitative research into a pilots attention allocation mechanism is required in the optimization design of an aircraft human-machine interface and system evaluation.After making a comprehensive consideration of several factors，including the importance of information， information detective efficiency and human errors，a pilot attention allocation model was built on the basis of hybrid entropy. To verify this model，a simulation model of a head-up display used to present flight indicators was developed. The experiments on the key-press response and eye-movement tracking were designed and carried out under the cruise and hold modes.As the experiment results are in good agreement with the theoretical model， the effectiveness of the pilot attention allocation model based on fuzzy theory is confirmed.However，the former model required to quantify the information detection efficiency which reduced its prediction ability.A new forecast model of attention allocation is established on the basis of information priority using fuzzy mathematics.The new model was built on four trade-off factors，including information priority，occurring probability， salient and effort， which involved both channels of information processing.To validate this model，sixteen participants were recruited to perform an instrument monitoring task under different conditions.The participants were required to concentrate on the flight indicators presented on the simulation interface of head-up display and respond to abnormal information by pressing the corresponding button on the keyboard. The regression analysis revealed good agreement between fixation distribution and model prediction results.

Key Words：Visual attention allocation；Mathematical modeling；Ergonomics experiment；Cognitive characteristics

閱讀全文鏈接（需實名注冊）：http：//www.nstrs.cn/xiangxiBG.aspx？id=51151&flag=1