語音訊號應用於中醫過敏性鼻炎病患之研究

A Study of Auscultical Signal in Traditional Chinese Medicine for the Allergic Rhinitis Patients

黃耀德(Yao-Te Huang)

指導教授 ： 翁清松

中原大學/工學院/醫學工程研究所/碩士(2002年)

https://doi.org/10.6840/cycu200200563

摘要

過敏性鼻炎為呼吸道的疾病之一，由於呼吸系統與發音系統確實有所關聯，所以在本研究中希望能從中醫聞診的角度來探討，以數位語音訊號處理的技術為工具，進而分析過敏性鼻炎患者聲音與健康人間的差異。
在本研究中，分析共有70名過敏性鼻炎病患（男成人20名，女成人50名）的語音資料，其資料由台北市立中醫醫院門診部所獲得。而健康人共有30名（男成人15名，女成人15名），成人資料則取自中原大學大學部及研究所健康的學生。受測者以平常發音之音調及音量，依序發音（[a]、[i]、[u]、[e]、[o]、[m]），發音前緩緩吸一口氣，然後開始發音一直到無法發音為止，其語音資料收錄在可攜式的迷你光碟機中，繼而將資料經由電腦音效卡送至電腦進行分析。
而所使用語音分析的參數，主要包含音高、越零率、共振峰、倒頻譜參數四方面。
實驗結果顯示，在病患和正常人之間有些語音有顯著的差異（p＜0.05）。男性方面，在[a]、[i]、[u]、[o]、[m]的越零率和[i]的第二共振峰、[u]的第三共振峰、[e]的第二共振峰、[o]的二、三、四共振峰、[m]的第二、四共振峰有差異。在女性方面，只有[a]的第二共振峰、[u]的音高，[e]的第一共振峰、[o]的第一、第二共振峰有差異性。
以 音高、越零率、共振峰分析，本系統之準確率及系統的信賴度Kappa值各為男性（準確率﹦82%，Ｋ﹦0.81），女性（準確率﹦80%，Ｋ﹦0.79），若增加倒頻譜參數分析，則男性（準確率﹦88%，Ｋ﹦0.87），女性（準確率﹦83%，Ｋ﹦0.82）。再者，若只增加倒頻譜參數，不包含鼻音[m]的分析，則男性為（準確率﹦85%，Ｋ﹦0.84）。

由本實驗結果可知，關於本過敏性鼻炎病症語音診斷系統，增加倒頻譜參數以及鼻音樣本[m]之語音分析，對於過敏性鼻炎病症的辨識率有明顯提升的效果。

關鍵字

共振峰 ； 中醫聞診 ； 倒頻譜參數 ； 越零率 ； 音高 ； 訊號處理 ； 過敏性鼻炎 ； 語音

並列摘要

Abstract
Allergic Rhinitis is one of the diseases of the respiratory tract. Since the respiratory system and the pronunciation system really have some relations, this research was to study Allergic Rhinitis in traditional Chinese medicine by the acoustical method. The technique of digital speech signal processing was used to analyze the difference between the voice of Allergic Rhinitis patients and that of the healthy ones.
In this research, the speech data of 70 Allergic Rhinitis patients (20 male adults, and 50 female adults) were acquired from Taipei Municipal Hospital of Traditional Chinese Medicine. The speech data of 30 healthy people (15 male adults, and 15 female adults) were collected. Adult’s speech data were from the healthy students in Chung-Yuan Christian University. The recording started on a portable mini optical disk when the sounds ( [a], [i], [u], [e], [o], [m]) were pronounced in normal pitch and volume by the subject. Before pronouncing, people relaxed and took a deep breath, and then started the pronunciation as long as possible. Finally, the speech data was acquired through the SoundBlaster card, and analyzed on the computer.
The parameters of the speech signals are: Pitch, Zero Crossing Rate(ZCR), Formant, Cepstrum Coefficient. The results showed that there were significant differences (p < 0.05) in sounds between patients and healthy people. In male, it has some varieties in the sounds [a] (ZCR)、[i] (ZCR, 2nd Formant)、[u]( ZCR, 3rd Formant)、[e](2nd Formant)、[o](2nd, 3rd, 4th Formant)、[m](2nd, 4th Formant). In female, the varieties of the sounds are [a](2nd Formant)、[u]（pitch）、[e](1st Formant)、[o](1st, 2nd Formant). According to the results, there were different voice parameters between healthy people and patients.
With the speech analysis of the pitch, zcr, and the formant, the results of the accuracy (A) and the kappa (k) value to the diagnosis system were：(Male：A=82%, K=0.81；Female：A=80%, K=0.79).When it included the cepstrum coefficient to analyze the voice, the results were：(Male：A=88%, K=0.87；Female：A=83%, K=0.82). If it included the cepstrum coefficient and excluded the nasal sample ([m] sound) to analyze the sound, the results to the male were (A=85%, K=0.84).
The results showed that adding cepstrum coefficienct and nasal sound [m] could increase the differenciation rate of the Allergic Rhinits diagnosis system.

並列關鍵字

pitch ； Allergic Rhinitis ； speech ； zero crossing rate ； digital processing for speech signals ； cepstrum coefficients ； Auscultation ； formant

參考文獻

【33】 陳修宜 徐茂銘：過敏性鼻炎的治療原則，當代醫學，28:10=336 pp 9-14，2001。

【1】 Alan V. Oppenheim: “Digital Processing of Speech”, Applications of Digital Signal Processing, pp 117-119, 1978.

【2】 Lawrence R. Rabiner, et al.: “A Comparative Performance Study of Several Pitch Detection Algorithms”, IEEE Trans. Acoust., Speech, and Signal Process., vol. ASSP-24, pp 399-417, Oct.,1976.

【14】 G. Feng & E. Castelli, “Some acoustic features of nasal and nasalized vowels: a target for vowel nasalization,” J. Acoust. Soc. Am,99 (6):3694-3706, 1996.

【15】 L. Ohde & A. Ochs, “Analysis of Nasal Consonants”, J. Acoust. Soc. Am, 34(12): 1865-1870, 1996.