語言學習與口語矯治對於聽語障礙者之早期療育扮演重要的角色，對於低年級兒童之語言發展遲緩難以診斷出來。西方國家所研發之語言資料庫無法直接轉移給本土語言，使得本土語言發展之評估僅有少數具信度與效度驗證的口語語言量表，嚴重影響語言教育、語言治療與語音科技發展。

本研究之特定目標為(1)利用以錄成完成之六到八歲低年級學齡兒童雙字詞分析擷取語音相關之基頻與第一共振峰(F1)加以量化；(2)將所擷取的六至八歲低年級學齡兒童之基頻與共振峰分析其不同年齡層與性別之間再基頻與第一共振峰的相關性或差異性，以作為建立母音常模資料庫與母音參數量化之基礎。本研究之分析方法為：(1)將所錄製完成之雙字詞分割成單字；(2)以音框分割方式將從單字擷取之母音ㄨ、ㄚ、一加以量化；(3)將已量化之母音ㄨ、ㄚ、一擷取其基頻與第一共振峰(F1)語音特徵參數；(4)將所擷取之基頻與第一共振峰(F1)經由統計方式來加以分析在不同年齡層與性別之間的關係。

本研究在語音處理方面決定兩個參數：(1)母音擷取臨界值：設定在單字最大音框能量之0.7倍以上為稱為母音；(2)AR model階數：設定AR model階數為20階，使在擷取第一共振峰時可得到較少之變異性與較高之一致性。在基頻與第一共振峰參數結果分析中，(1)基頻方面：不會因不同的年齡層與性別而有所差異；(2)第一共振峰(F1)：會因性別的不同而有所差異，女生之第一共振峰高於男生之第一共振峰；不會因不同的年齡層而有所差異。

本研究之目的於針對台灣六至八歲學齡兒童，將所發出之語音訊號加以分析，分析語音參數與年齡和性別之間是否相關，做為母音分類分群之基礎，進而提供相關母音量化之依據，對於特教語言教學評估工具、教具訓練之教材、計算語言學之研究，提供系統化與科學化之語言語音發展。

Language learning and speech therapy are the important issue in early intervention for language disorders. Language development delay for school children is hard to diagnose. Developed countries have developed lexicon database, which can’t be transferred and applied into our native language lexicon database as Mandarin and Taiwanese. Assessments of local language development for verbal language scale are only a few reliable and valid lexicon databases. Lack of domestic database is detrimental to language education, training, therapy and speech technology.
The purpose of this research is to analysis speech signal and quantitative pitch & F1 from 6 to 8 school children. The specific aims of this research study are (1) to extract and quantify pitch frequency and F1 formant frequency from recorded completely bi-syllables speech for 6 to 8 years old school children, (2) to analyze and compare correlation and diverseness of pitch and F1 in order to build vowel normal database and basis of quantitative vowel parameter. Research methods include：(1) to segment two word into one bi-syllables word, (2) to separate bi-syllables words and extract vowel ㄨ/u/、ㄚ/a/、一/i/ quantitatively, (3) to analyze the effect of age and sex on pitch and F1 for ㄨ/u/、ㄚ/a/、一/i/ vowel .
In the speech analysis, two impartment speech process parameters：(1) vowel extraction threshold value setting exceeds maximum frame energy multiply by 0.7 is determined, (2) the order number of AR model is determined for 20 to minimize the variation and optimize the consistency. The results show that age has statistically no effect on the pitch of ㄨ/u/、ㄚ/a/、一/i/ vowels (p>0.1), however, female children have higher pitches of ㄨ/u/、ㄚ/a/、一/i/ vowels than those of male children(P<0.05).The results show that age has statistically no effect on the F1 of ㄨ/u/、ㄚ/a/、一/i/ vowels (p>0.1), however, female children have higher F1 of ㄨ/u/、ㄚ/a/、一/i/ vowels than those of male children(P<0.05).
This research contributions may be signification to assessment of special language education、clinical speech therapy and training、calculation speech technology as well as application to domestic language and speech development.

1.佘永吉, 台灣學齡前兒童口與詞彙資料庫之發展, 2001
2.何國華, 特殊兒童心理與教育, 五南圖書出版公司, 1987
3.洪維恩, MatLab 7 程式設計, 旗標出版股份有限公司, 2005
4.羅華強, 訊號處理—MATLAB的應用, 全華科技圖書股份有限公司, 2003
5.王小川, 語音訊號處理, 全華科技圖書股份有限公司, 2007
6.杜琬之, 言語科學概論—從基礎理論到臨床應用, 心理出版社, 2005
7.謝依蘭, 語音訊號數位處理, 松崗電腦圖書資料股份有限公司, 1992
8.楊鎮光, Visual Basic 與語音辨識—讓電腦聽話, 文魁資訊圖書資料股份有限公司, 2002
9.臺灣師範大學國音教材編輯委員會, 國音學, 正中出版社, 2007
10.王保進, 視窗版SPSS與行為科學研究, 心理出版社, 1999
11.SPSS for Windows--統計分析-初等統計與高等統計, 松崗電腦圖書資料股份有限公司, 2004
12.陳順宇, 統計學, 華泰書局, 1996
13.陳順宇, 多變量分析, 華泰書局, 2004
14.林永松, 張簡培崙, 沈美玲, 中文揚揚格語詞聽閥檢定表. 耳鼻喉科學會雜誌, 1997. 32(1): p.7-13
15.Simon Haykin, Barry Van Veen, Signal and System 2nd, John Wiley & Sons, INC. ,2002
16.Furui, Sadaoki, Digital speech process, Synthesis, and Recognition, TOKAI University Press, 1985
17.Alan V. Oppenheim, Digital processing of speech signals, Bell Laboratories,INC. , 1978
18.Thomas F. Quatieri, Discrete-time speech signal processing — principles and practice, MIT Lincoln Laboratory, 2002
19.Lawrence Rabiner, Biing-Hwang Juang, Fundamentals of speech recognition, Prentice-Hall INC.,1993
20.Huang, Xuedong ect., Spoken language processing, Prentice Hall PTR, 2001