耳鳴是大部份的人都曾經有過的經驗，在老年人口中更高達三分之一有耳鳴的問題。根據美國耳鳴協會的資料顯示，在全美約有四千四百萬的耳鳴患者；其中一千兩百萬的患者為嚴重的耳鳴所困擾。不幸的是，傳統的醫學治療方式卻無法消除大多數耳鳴患者的狀況，許多患者覺得被放棄、困惑，並開始怨天尤人。這種情形持續到1980年，神經生理學學者Pawel Jastreboff及Dr.Vernon分別提出了Tinnitus Retraining Therapy及Tinnitus Masking的耳鳴治療方法，這兩種方法都包含了專科醫師的臨床諮商及使用耳鳴治療器的部份。其中耳鳴治療器的功能是基於神經的可塑性、適應性等生理特性以加速患者適應耳鳴並進而改善耳鳴的症狀，達到復健及治療的目的。
　　然而目前巿面上歐美進口的耳鳴治療器主要以寬頻噪音為治療音源，忽略在耳鳴患者中有約七成的比例同時併有聽力損失的問題。使用僅有音量調整功能的白色噪音耳鳴治療器不但使得治療期程冗長、對部份患者效果不彰且潛在著傷害正常聽力頻段的隱憂。價格昂貴也使得許多有需要的耳鳴患者望之卻步，失去獲得有效治療機會。
　　近年來數位訊號處理的技術在軟硬體方面都有顯著的進步，使得實現複雜的訊號處理演算法變成可能。本研究結合了Tinnitus Retraining Therapy及Tinnitus Masking兩種方法的特點，應用了統計數位信號處理及模型的原理建構一套PC-based的耳鳴量測暨復健平台，藉由在平台上對患者實施聽力量測以獲得其聽力狀態及耳鳴頻段音量等參數，根據這些聽力生理參數在平台上迅速產生具有特定頻率響應的復健治療用雜訊。
本研究的耳鳴量測暨復健平台目前已建構完成，其中包括純音閥值（PTA）模組介面、窄頻雜訊模組介面、耳鳴音逼近模組介面及特定頻譜雜訊產生模組介面共四個部份。純音閥值模組提供250、500、1 k、2 k、4 k及8 k Hz的純音訊號以量測受測者聽力閥值。窄頻雜訊模組則提供從20∼20k Hz分成18個頻段的窄頻雜訊以匹配耳鳴音頻段與最小遮蔽值（MML）、混合點（MP）音量參數。耳鳴音逼近模組提供可彈性調整頻寬之雜訊以彌補窄頻雜訊模組的固定頻段限制。特定頻譜雜訊產生模組則根據量測所得聽力參數產生指定頻譜之耳鳴復健治療雜訊。在平台系統的校正方面，頻率校正結果顯示相對誤差值都在0.1%以下。而平台設定音量與實際輸出音量的對照表也已完成，平台輸出的純音或雜訊訊號，皆可由對照表得到實際輸出音量值（dB SPL）。在線性度方面，標定值與實際輸出音量的關聯係數之平均值為0.9724±.03445。整個耳鳴量測暨復健平台系統可產生頻率範圍介於20∼20 kHz，音量最高達117.58 dB SPL的雜訊。
目前並完成初步階段的耳鳴臨床量測測試。所收集的21位耳鳴患者的測試結果中，聽力曲線呈現高頻聽損的患者佔最多數（7/21）。耳鳴位置以左側居多（10/21）。患者的耳鳴音頻率分佈並無特別的趨勢。在耳鳴音頻率匹配方面可達95％（20/21）；而在音量匹配有75％（15/20）。MML的平均音量值為97.27±18.8 dB ，而MP的平均音量值為86.18±16.7 dB。兩者間的迴歸方程式為y=.839x+7.266，可以MML值預測MP值。從目前初步測試階段的結果及受測者的反應看來，雖然仍有改善空間，但整體上對大部份的患者而言，平台在量測方面的效果尚稱理想。
本平台系統未來可朝下列方向持續研究及發展：(1)將平台所產生的耳鳴治療復健雜訊下載至MP3播放器作為耳鳴治療器，建立使用本土化耳鳴治療器之實驗組及使用歐美進口耳鳴治療器之控制組，進行試戴實驗及療效比較分析，(2)可考慮應用平台所產生的線性預測參數組，以DSP晶片系統硬體化實現本土化耳鳴治療器。

Most people have tinnitus experiences with various degree, and one-third of the elderly are suffered from tinnitus. The statistical data of American Tinnitus Association shows that there are forty-one million tinnitus patients in America. Twelve million patients have severe tinnitus. Unfortunately, most conventional treatments fail to improve the symptoms of tinnitus. Many patients feel aborted, confused and depressed. Until 1980, neurophysiologist Pawel Jastreboff and Dr. Vernon have proposed a Tinnitus Retraining Therapy and a Tinnitus Masking treatment for the tinnitus, respectively. These therapies consist of clinical consultation by physicians and applying noise generator for rehabilitation treatment. Based on the plasticity and habituation of neuron, the implementation of noise generator is to accelerate the period of habituation to improve the symptoms of tinnitus.
Most imported commercial noise generators provide primarily broadband noises for the treatment ; unfortunately, the devices and their corresponding treatment are without considering the clinical evidence that among 70% of tinnitus patients are complicated with hearing loss. Except white noise with only volume adjustment from noise generators providing inefficacious outcomes for some tinnitus patients, it always takes a prolong period for treatment and even is at high risk for hearing loss damage. Moreover, the imported noise generators are too expensive for many domestic patients in need to obtain medicinal treatment.
Recently, Advanced technology of digital signal processing in both hardware makes the implementation of many complex algorithms becoming possible. Based on the concept and underlying mechanisms of Tinnitus Retraining Therapy and Tinnitus Masking, this research project, which applies theories of statistical digital signal processing and modeling, is aimed to develop and establish a PC-based platform for evaluation and rehabilitation of tinnitus. This platform, which provides the measuring and evaluation of physical parameters of hearing state and tinnitus, is able to prescribe the customized noise of treatment with specified spectrum.
The evaluation and rehabilitation platform for tinnitus treatment are completely developed with four main modules including PTA module, narrow-band noise production module, tinnitus-sound approaching module, and noise with specified spectrum production module. There are pure tones of 250, 500, 1k, 2k, 4k and 8k Hz in the PTA module to measure subject’s hearing thresholds. In narrow-band noise production module, it provides 18 sub-bands covering from 20 to 20k Hz of noise to evaluate the frequency band ,minimum masking level(MML) and mixing point(MP) of tinnitus. The tinnitus-sound approaching module offers noise with adjustable bandwidth to resolve the limitation of fixed band of narrow-band noise production module. The noise with specified spectrum production module produces the noise of tinnitus rehabilitation and treatment with specified spectrum according to the measured hearing parameters.
The calibration and testing of the platform system has been conducted . The results of frequency calibration of pure tones indicate that all the relative errors for 250, 500, 1k, 2k, 4k and 8k Hz are less than 0.1%. The mapping table of setup volume in this system to sound level meter output is also conducted. Output volume of pure tones or noise of the platform could be converted to be in dB SPL by this mapping table. The calibration results of volume output of the system relative to sound level meter show that the mean value of correlation coefficients between nominal value and system output volume is 0.9724±.3445. This PC-based platform could provide the noise with bandwidth ranging from 20 to 20k Hz and volume up to 117.58 dB SPL for evaluation and rehabilitation of tinnitus.
A pilot study is conducted to clinically implement the system for measuring and evaluation of tinnitus patients. A total of 21 tinnitus patients are in involved at this stage. For the PTA measurements, 7 of 21 patients have show high frequency loss type. 10 of 21 patients are Left-side tinnitus . There is no obvious trend in the distribution of tinnitus frequency, it might be resulted from small size of sampling. Using this platform 20 of 21 patients are frequency matching in tinnitus, and 15 of 20 patients are volume matching. The mean and standard deviation of volume values for MML and MP are 97.27±18.8 dB and 86.18±16.7 dB ,respectively. This system provides the prediction capability for the value of MP through regression analysis of MML and MP with equation y=.839x+7.266 (x : MML; y : MP), correlation coefficient=0.972 at p<.05. This platform works well so far by the results of clinic trial stage and responses of subjects.
The future system research and development are recommended as following: (1) to download the customized noise to MP3 player as a noise generator and to setup the experimental group with customized-noise generator and control group with imported white noise generator for the comparison and analysis of the results of both methods; and (2) to implement the noise generator by DSP chip with linear prediction coefficients calculated by this platform system.

中文摘要 I
Abstract Ⅲ
誌謝 Ⅵ
目錄 Ⅶ
表目錄 XI
圖目錄 XⅡ
第一章 緒論 1
1.1 前言 1
1.2 背景及文獻回顧 2
1.2.1 聽力系統的解剖生理 2
　　　　　聽力解剖構造 2
　　　　　聽力生理機轉 3
1.2.2 耳鳴的發生及病因 5
　　　　　耳鳴理論：耳鳴如何發生 5
　　　　　耳鳴的病因 6
1.2.3 耳鳴治療方式 7
　　　　　耳鳴遮蔽法 ( Tinnitus Masking ) 7
　　　　　耳鳴減敏療法 ( Tinnitus Retraining Therapy ) 8
1.2.4　耳鳴患者的類型 9
1.2.5　耳鳴治療用之雜訊產生方式 11
1.2.6　耳鳴治療雜訊訊號處理之數位式濾波器設計 13
　　　　　數位式濾波器的種類 14
　　　　　FIR濾波器的設計方式 16
　　　　　視窗法（Window Method） 16
　　　　　最佳法（Optimal Method） 19
　　　　　頻率取樣法（Frequency Sampling Method） 21
　　　　　 非遞迴式頻率取樣濾波器 21
　　　　　 遞迴式頻率取樣濾波器 23
1.3 研究動機、目的及特定目標 25
1.4 研究假說及重要性 26
Research Hypothesis 26
Significances 27
第二章 系統設計與材料方法 28
2.1 系統設計理念 28
2.2 系統功能 29
2.3 耳鳴量測復健平台及各模組介面之設計 30
純音閥值量測模組介面 32
窄頻雜訊音產生模組介面 34
近似耳鳴音產生模組介面 40
復健音產生模組介面 44
2.4 平台校正程序 47
頻率校正 47
音量校正 48
2.5 研究實驗流程 49
資料分析及病患之基本資料庫建立 51
第三章 結果與討論 53
3.1 白色雜訊序列的特性 53
3.2 耳鳴量測及復健平台系統之功能特性 56
3.3 平台系統純音校正 62
頻率校正 62
音量校正 63
3.4 研究數據 67
3.5 耳鳴音匹配結果與討論 75
第四章 結論與未來展望 77
4.1 結論 77
4.2 未來展望 77
參考文獻 79
附錄A　平台純音輸出頻率量測值 81
附錄B　平台純音輸出音量量測值 82
附錄C　窄頻段雜訊輸出音量量測值 85

【1】 蕭雅文, 聽力學導論, 五南, 1997.
【2】 Abraham Shulman, Tinnitus diagnosis/treatment, Lea & Febiger, 1991.
【3】 A.Rodriguez-Vazquez,M. Delgado, S. Espejo, and J. L. Huertas, “Switch-capacitor broadband noise generator for CMOS VLSI”, Electronics Letters,vol. 27, pp.1913-1915, Oct. 1991.
【4】 A.Rodriguez-Vazquez,M. and Delgado, S. Espejo, “Nonlinear switched-current CMOS IC for random signal generation”, Electronics Letters,vol. 29, pp.2190-2191, Dec. 1993.
【5】 Bernhard Kellerhals and Regula Zogg, Tinnitus Rehabilitation by Retraining, Karger, 1999.
【6】 Craig S. Petrie, et al., “A Noise-Based IC Random Number Generator ” IEEE, 2000.
【7】 Coles RRA, “Epidemiology of tinnitus: (1) Prevalence”. J Laryngol Otol, 91(Suppl.): 7-15, 1984.
【8】 Creating Graphical User Interfaces, Version 6, MathWorks, MA, 2002.
【9】 Database Toolbox User’s Guide, Version 2, MathWorks, MA, 2002.
【10】 David W. Rees and Simon D. Smith, Living with tinnitus, Manchester University, 1991.
【11】 Drukier GS, “The prevalence and characteristics of tinnitus with profound sensori-neural hearing impairment”, Am Ann Deaf :134 260-264, 1989.
【12】 Emmanuel C. Ifeachor and Barrie W.Jervis, Digital Signal Processing, Addison-Wesley, 1993.
【13】 Eugene N. Bruce, Biomedical Signal Processing and Signal Modeling, John Wiley & Sons, New York, 2001.
【14】 Filter Design Toolbox User’s Guide, Version 2, MathWorks, MA, 2002.
【15】 Hazell JW, Jastreboff PJ, Tinnitus. I: ”Auditory mechanisms: a model for tinnitus and hearing impairment”, J Otolaryngol 19(1):1-5. ,1990.
【16】 Http://www.ata.org/. American Tinnitus Association ,2002.
【17】 Http://www.xilinx.com/. Xilinx Application Notes, 2002.
【18】 Jastreboff PJ et al. Neurophysiological approach to tinnitus patients, Am.J.Otology1996; 10:464-472. ,.
【19】 Jastreboff PJ et al. Optimal sound use in TRT-theory and practice. Proceedings of Sixth International Tinnitus Seminar, 1999, Cambridge, UK. London, UK.
【20】 Jastreboff MM et al. Questionnaires for assessment of the patients and treatment outcome. Proceedings of Sixth International Tinnitus Seminar, 1999, Cambridge, UK. London, UK.
【21】 Jastreboff PJ, Sasaki CT. An animal model of tinnitus: a decade of development; Am J Otol1994; 15(1):19-27.
【22】 Kauer JS, Nemitz JW, Sasaki CT. Tinnitus aurium: fact…or fancy. Laryngoscope1982; 92:1401-1407.
【23】 Leske M, “Prevalence estimates of communicative disorders in the U.S.: Language,learning and vestibular disorders”, ASHA 23:229-237, 1981.
【24】 Monson H. Hayes, “Statistical Digital Signal Processing and Modeling”, John Wiley & Sons, New York, 2000.
【25】 Pellerin G, Magistretti PM, “Glutamate uptake into astrocyte stimulates aerobic glycolysis: a mechanism coupling neuronal activity to glucose utilization”, PNAS: 91: 10625-10629, 1994.
【26】 Richard S. Tyler, Tinnitus Handbook, Singular, 2000.
【27】 Sadaoki Furui, Digital Speech Processing, Synthesis, and Recognition, 開發, 1989.
【28】 Signal Processing Toolbox User’s Guide, Version 6, MathWorks, MA, 2002.
【29】 Searchfield GD et al. “Digital instruments for tinnitus: mixing point identification and threshold-adjusted noise”, Seventh International Tinnitus Seminar, Fremantle, Australia, 2002.
【30】 Sedra Smith, Microelectronic Circuits, OXFORD, 4th edition, New York, 2002.
【32】 Sheldrake JB et al. Results of tinnitus retraining therapy. Sixth International Tinnitus Seminar, Cambridge, UK, 1999.
【33】 W. Timothy Holman et al. ,“ An Integrated Analog/Digital Random Noise Source”, IEEE, 1997.