臺灣博碩士論文加值系統

生活中常聽到疑似表裡不一的詞語，像是朋友變胖了，卻要當面問我的看法。為了避免對方難過，只好說體態維持很好。這些話自己聽起來覺得心虛，與平常說話的語調有差異，卻又無法具體說明差別在哪。能否發展出一種工具來輔助我們判斷說話者語調透露出的訊息? 為此，提出一個表裡不一實驗，分為Baseline和Deception兩階段。Baseline會問受測者一些簡單問題，如早餐吃什麼，其目的為蒐集心情平靜且真誠無隱瞞的聲音。Deception階段要求受測者選出十張不喜歡的圖片，接著假裝很喜歡這些圖片，且要說出喜歡的原因。全程採用中文來提問與答覆，並且錄音，再將錄音檔切割出所對應的聲音區塊加以分析。錄音檔用Praat軟體將16個母音(ㄚㄛㄜㄝㄞㄟㄠㄡㄢㄣㄤㄥㄦㄧㄨㄩ)進行JTFA轉換以獲得個別母音的f0、f1與f2，並請一位執業14年的心理諮商師重新聽二階段的聲音，採五分法來判讀表裡不一的程度。在交大REC中心的同意下共招募30人參與此實驗，結果呈現，(1) 以“ㄣㄟㄜ”之(f1,f2)為座標所圍成之向量面積，在30人當中有25人(83%)出現Dec_area > Bas_area，統計檢定結果為p= .01，可當成識別「表裡不一」的指標；(2)若在不考量樣本數量的條件下，則研究所p = 0.000015(11人)；大學p = 0.000083(12人)；專科p = 0.0074(5人)；國、高中p = 0.21(2人)。隨著學歷下降，p-value有逐漸上升的趨勢，統計檢定顯示，隨著學歷越高，對說謊特徵的掌握度就越好。

We often hear some deceptive speechs in our lives. Just like a friend looks fat, but when he asked me about how fat he is, I pretended not to think so. When we lie, the tone of these speech sounds strange but we do not unaware ourself. Is it possible to develop a tool for helping us to judge the speaker's insight thoughts? To this end, designed a deceptive experiment. The experiment is divided into two stages: Baseline and Deception. Baseline stage will ask the subject some simple questions, such as what to eat at breakfast, the purpose is to collect non-deceptive voice. The Deception stage requires the subjects to select ten pictures they do not like, and then pretend to like the pictures, and say why they like them. The course of the experiment records speaker’s speech and PPG information. JTFA is performed with the Praat software to obtain f0, f1 and f2 of the 16 vowels (ㄚㄛㄜㄝㄞㄟㄠㄡㄢㄣㄤㄥㄦㄧㄨㄩ). And find a psychological counselor (14 years of qualification) to listen to the sound of two stages to determine the extent of Deception in 1 to 5 points. The results show, (1) 25 of the 30 people showed a greater trend in Dec_area (f1,f2) vector area under“ㄣㄟㄜ”,and there were statistically significant differences (p=.01). The (f1, f2) vector area can be used as a criterion for identifying " Deception "; (2) If we do not consider the sample size, the higher the degree of education, the higher the ability to lie.

摘要 …………………………………………………………………………………………………………………………… i
Abstract ……………………………………………………………………………………………..……………………………… ii
誌謝 ………………………………………………………………………...…………………………….…… iii
目錄 ………………………………………………………………………….………………………….…… iv
表目錄 ……………………………………………………………………………………..…………………...………… v
圖目錄 …………………………………………………………………………………….………………………….…… vi
一、 緒論…………………………..……………………………………………………..……………………….… 1
1.1 背景…………………………………………………………………………………………………..……….…1
1.2 動機……………………………………………………………………………………………………..……….1
1.3 五種知覺與大腦對應位.………………………………………………………………...……2
1.4 人類傳遞情緒的方法……..…………………………………………………………………………….7
1.5 音頻分析文獻……………………………………………………………………………………………..11
1.6 目的…………………………………………………………………………………………………………….15
二、 研究內容與方法……………………………………………………….………………………………..16
2.1 實驗設計……………………………………………………………………….…………………………….16
2.2 語音處理流程……………………………………………………………………………………………..19
三、 實驗結果……………………………………………………………………………………………….......23
3.1 實驗資料獲取………………..…………………………………………………………………… ……..23
3.2 實驗數據………………………..……………………………………………………………………………24
3.3 (f1,f2)多邊形向量面積…………………….…………………………………………………………..41
3.4 心理諮商師評分結果………………………………………………………………………………….42
四、 討論…………………………………….……..……………………………………………………………… 43
4.1 母音面積與心理諮商師評分..…………………………………………………………………… 43
4.2 指標f0之討論………………………………………………………………………………………….… 47
4.3 指標f1之討論……………….………………………………………………………………………….…48
4.4 指標f2之討論……………….………………………………………………………………………….…49
4.5 多邊形之(f1,f2)向量面積討論………………..……………………………………………………50
五、 結論…………………………………………..………………………………………………………………..51
六、 未來工作…………………………………..………………………………………………………………..52
參考文獻 …………………………………………………………………………………………………………………….53
附錄一 …………………………………………………………………………………………………………………….54

[1] 周德程，解剖生理學，昭人出版社，台北，民國六十一年。
[2] （美）William F. Ganong著，GANONG醫學生理學，賴義隆譯，麥格羅希爾，台
北，民國九十七年。
[3] HEILVEIL, I., "Deception and pupil size", Journal of Clinical Psychology, vol. 32, no. 3,
pp. 675-676, July 1976.
[4] Lippold, O., "Physiological tremor", Scientific American, vol. 224, pp. 65-73, 1971.
[5] Clifford S. Hopkins, Daniel S. Benincasa, Roy J. Ratley, John J. Grieco, “Evaluation of voice stress analysis technology”, Hawaii International Conference on System Sciences (IEEE), USA, Jan 2005.
[6] Liu,Xianfeng, “Voice stress analysis: Detection of deception”, Univ. of Sheffield, MSc thesis, 2005.
[7] Cosetl, R. C. ,et al., “Voice stress detection: A method for stress analysis. Detecting fluctuations on Lippold Microtremor spectrum using FFT”, pp. 184-189, 21st International Conference on Electrical Communications and Computers (IEEE), Mexico, April 2011.
[8] Kirchhübel. C, “The acoustic and temporal characteristics of deceptive speech”, Univ. of York, PhD thesis, 2013.
[9] Dektor Corporation, “PSE7010”, http://www.dektorpse.com/, 2017.
[10] NITV, “CVSAII”, https://www.cvsa1.com/, 2017.
[11] Nemesysco, “LVA6.50”, http://nemesysco.com/, 2017.
[12] Paul Boersma and David Weenink, “Praat: doing phonetics by computer”, http://www.praat.org/, 2017.
[13] 行政院國家科學委員會, “Taiwan Affective Picture System”, http://ssnre.psy.ntu.edu.tw, 2017.
[14] Paul Ekman, “An argument for basic emotions”, Cognition and Emotion, pp. 169-200,
1992.
[15] Carl Fridrich Gauss, “Gauss's area formula ”, 1795.