【摘要】：認(rèn)知負(fù)荷是人在處理信息過程中所消耗的認(rèn)知資源。由于認(rèn)知負(fù)荷水平會顯著影響人類執(zhí)行任務(wù)的效率,所以在教育方法改進(jìn)、交互產(chǎn)品設(shè)計(jì)、高壓力工作監(jiān)測等領(lǐng)域,都需要對認(rèn)知負(fù)荷進(jìn)行科學(xué)的測量。目前測量認(rèn)知負(fù)荷的方法可分為三類,分別是主觀評定測量、任務(wù)績效測量和生理測量。主觀評定測量通過人的主觀感受和體驗(yàn)評估認(rèn)知負(fù)荷,需要依托于評價(jià)量表,容易受到主觀影響。任務(wù)績效測量根據(jù)任務(wù)中的績效成績間接評估認(rèn)知負(fù)荷,其績效成績?nèi)菀琢炕徒y(tǒng)計(jì),但其指標(biāo)必須根據(jù)任務(wù)而設(shè)定。生理測量是一種客觀、可量化的測量方法,其中眼動數(shù)據(jù)可以以非接觸的方式采集,具有較高的實(shí)際應(yīng)用價(jià)值。所以本文基于認(rèn)知負(fù)荷理論設(shè)計(jì)了認(rèn)知負(fù)荷眼動數(shù)據(jù)采集實(shí)驗(yàn),對能夠體現(xiàn)認(rèn)知負(fù)荷的眼動特征進(jìn)行了分析,提出了一種剔除用戶個(gè)體差異的特征分析方法,并結(jié)合模式識別的理論方法完成了對認(rèn)知負(fù)荷水平的測量。本研究分為兩個(gè)階段,其主要研究內(nèi)容如下:第一階段實(shí)現(xiàn)對兩種認(rèn)知負(fù)荷狀態(tài)的識別。采用判斷任務(wù)的實(shí)驗(yàn)范式誘發(fā)認(rèn)知負(fù)荷;借助統(tǒng)計(jì)檢驗(yàn)確定了12個(gè)能夠體現(xiàn)認(rèn)知負(fù)荷狀態(tài)的特征;提出一種去除眼動特征中被試間差異的方法;利用支持向量機(jī)(SVM)完成對認(rèn)知負(fù)荷狀態(tài)的識別,其識別準(zhǔn)確度為90.2%;根據(jù)識別結(jié)果確定認(rèn)知負(fù)荷狀態(tài)的最優(yōu)特征。第二階段實(shí)現(xiàn)對認(rèn)知負(fù)荷水平的量化。采用心算任務(wù)的實(shí)驗(yàn)范式,通過控制計(jì)算難度操縱認(rèn)知負(fù)荷水平;提取55個(gè)眼動特征,并詳細(xì)分析興趣區(qū)駐留時(shí)間占比、計(jì)算過程中瞳孔大小改變量與認(rèn)知負(fù)荷水平的關(guān)系;利用序列后向選擇算法(SBS)和支持向量機(jī),確定最優(yōu)的特征組合并完成對認(rèn)知負(fù)荷多個(gè)水平的識別,其識別準(zhǔn)確率為74.4%;借助于分類的后驗(yàn)概率,完成對認(rèn)知負(fù)荷的量化。本研究利用眼動數(shù)據(jù)識別認(rèn)知負(fù)荷狀態(tài)、水平,并進(jìn)一步完成對認(rèn)知負(fù)荷水平的量化,從而說明了借助于眼動數(shù)據(jù)測量認(rèn)知負(fù)荷水平的可行性。鑒于眼動數(shù)據(jù)采集的非接觸性,上述研究結(jié)果有望推廣到實(shí)際的應(yīng)用場景之中。
[Abstract]:Cognitive load is the cognitive resource consumed by people in the process of processing information. Because the cognitive load level can significantly affect the efficiency of human task execution, it is necessary to measure the cognitive load scientifically in the fields of educational method improvement, interactive product design, high stress job monitoring and so on. At present, the methods of measuring cognitive load can be divided into three categories: subjective evaluation, task performance measurement and physiological measurement. Subjective assessment measures assess cognitive load through people's subjective feelings and experiences, which depends on the evaluation scale and is susceptible to subjective influence. Task performance measurement indirectly evaluates the cognitive load according to the performance achievement in the task, its performance achievement is easy to quantify and statistics, but its index must be set according to the task. Physiological measurement is an objective and quantifiable measurement method, in which eye movement data can be collected in a non-contact manner, which has high practical application value. Therefore, based on the cognitive load theory, this paper designs the cognitive load eye movement data acquisition experiment, analyzes the eye movement characteristics which can reflect the cognitive load, and puts forward a feature analysis method to eliminate the individual differences of users. Combined with the theory and method of pattern recognition, the cognitive load level is measured. This study is divided into two stages. The main contents are as follows: the first stage realizes the recognition of two cognitive load states. The cognitive load was induced by the experimental paradigm of judgment task, 12 characteristics of cognitive load state were determined by statistical test, and a method to eliminate the differences in eye movement characteristics was proposed. The recognition accuracy of cognitive load state is 90.2 by using support vector machine (SVM), and the optimal feature of cognitive load state is determined according to the recognition result. In the second stage, the level of cognitive load is quantified. Using the experimental paradigm of mental arithmetic task, the cognitive load level is controlled by controlling the difficulty of calculation, 55 eye movement features are extracted, and the proportion of resident time in the area of interest is analyzed in detail, and the relationship between the pupil size change and the cognitive load level in the calculation process is analyzed in detail. Using (SBS) and support vector machine, the optimal feature combination is determined and the recognition of multiple levels of cognitive load is completed. The recognition accuracy is 74.4%. By means of the posteriori probability of classification, the quantification of cognitive load is completed. This study uses eye movement data to identify cognitive load state and level, and further completes the quantification of cognitive load level, which shows the feasibility of measuring cognitive load level with eye movement data. In view of the non-contact nature of eye movement data acquisition, the above results are expected to be extended to practical application scenarios.
【學(xué)位授予單位】：西南大學(xué)
【學(xué)位級別】：碩士
【學(xué)位授予年份】：2017
【分類號】：TP18;TP274

【參考文獻(xiàn)】

相關(guān)期刊論文 前10條

1 戢曉峰;馮川;郭鳳香;;ATIS環(huán)境下駕駛員認(rèn)知負(fù)荷研究進(jìn)展[J];安全與環(huán)境學(xué)報(bào);2015年03期

2 劉遠(yuǎn);劉洪廣;;基于眼動追蹤測量認(rèn)知負(fù)荷變化的犯罪心理測試[J];中國司法鑒定;2014年06期

3 張同柏;;認(rèn)知負(fù)荷理論研究:問題挑戰(zhàn)與融合超越[J];外國教育研究;2012年11期

4 任桂琴;韓玉昌;于澤;;句子語境中漢語詞匯形、音作用的眼動研究[J];心理學(xué)報(bào);2012年04期

5 白學(xué)軍;孟紅霞;王敬欣;田靜;臧傳麗;閆國利;;閱讀障礙兒童與其年齡和能力匹配兒童閱讀空格文本的注視位置效應(yīng)[J];心理學(xué)報(bào);2011年08期

6 周鵬生;周愛保;;THOG推理影響因素的眼動研究[J];西南大學(xué)學(xué)報(bào)(自然科學(xué)版);2011年02期

7 陳慶榮;譚頂良;鄧鑄;徐曉東;;句法預(yù)測對句子理解影響的眼動實(shí)驗(yàn)[J];心理學(xué)報(bào);2010年06期

8 李金波;許百華;;人機(jī)交互過程中認(rèn)知負(fù)荷的綜合測評方法[J];心理學(xué)報(bào);2009年01期

9 羅娜;;數(shù)據(jù)挖掘中的新方法——支持向量機(jī)[J];軟件導(dǎo)刊;2008年10期

10 陳巧芬;;認(rèn)知負(fù)荷理論及其發(fā)展[J];現(xiàn)代教育技術(shù);2007年09期

相關(guān)博士學(xué)位論文 前3條

1 孫崇勇;認(rèn)知負(fù)荷的測量及其在多媒體學(xué)習(xí)中的應(yīng)用[D];蘇州大學(xué);2012年

2 龔德英;多媒體學(xué)習(xí)中認(rèn)知負(fù)荷的優(yōu)化控制[D];西南大學(xué);2009年

3 彭曉武;VDT作業(yè)腦力勞動負(fù)荷評價(jià)的實(shí)驗(yàn)研究[D];華中科技大學(xué);2006年

，

本文編號：2333440