隨著台灣人口結構的老化及科技文明造成的嚴重噪音污染，導致不同類型及程度聽覺障礙的族群有日益增加的趨勢。根據統計顯示：台灣地區65~74歲中有18%具不同程度的聽力損失，75歲以上高達35%。根據美國NIDCD統計，大於65歲的老人中聽障人口約佔54%，實際配戴助聽器的人數僅佔聽力損失患者20％的比例，顯示出目前市面上的助聽器尚有不少地方值得改進，如語音辨識能力的加強、回音消除及自動音量增益控制等。本研究的目的為應用數位訊號處理及語音分析與合成技術發展一套聽力評估、聽力增益處方及治療訓練的PC-based系統平台。

　　助聽復健平台以MATLAB 6.5及Simulink/DSP blocket為開發工具。純音閾值量測模組，由正弦函數數列乘上振幅後輸出。增益處方匹配模組，應用speech analysis-synthesis 理論，將語音訊號以漢明窗音框化，然後以FIR濾波器組（低通濾波器、帶通濾波器及高通濾波器）將語音切割為10個次頻帶進行合成，輸出調整後增益的語音訊號。

　　本研究之助聽復健平台目前建構完成的部分包括純音閾值量測模組及增益處方匹配模組。純音閥值量測模組提供250、500、1 k、2 k、4 k、8 k及16kHz的純音訊號量測受測者聽力閥值。增益處方匹配模組提供從20~20 kHz 的語音等化器，音量最高達105dB SPL，將聲音分為10個語音次頻帶，配合增益調整策略，進行音量補償。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。平台設定音量與實際輸出音量的對照表也已完成。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0.976。

　　本研究平衡單字詞表建立部分，以中文語音平衡句為分析材料，建立音節出現次數統計表，挑選出50個平衡單字詞。以本研究挑選的單字詞為材料，進行語音理解評估測試，結果顯示正常聽力受測者語音理解度平均為92.5％，標準差為2.33％。受測人數僅有八名，測試結果尚稱理想。

　　本平台系統未來可朝下列方向持續研究及發展：（1）頻帶切割方面:運用多速率訊號處理，降低運算量提升效率；(2)消除噪音方面:加入適應性濾波器演算法，提升語音對噪音的訊噪比；（3）應用平台Simulink/DSP blockset的架構模擬測試，並將完成的程式轉譯為C code，由DSP板執行演算法的程式。

　　An increasing of the elderly and serious noise pollution on living environments have led more and more people to suffer from hearing impairments with various types and different degrees of hearing loss. In Taiwan, the prevalence of hearing impairments are reported that about 18 % of adults between the ages of 65 and 74 years have hearing loss ,and approximately 35% of the person over age 75 have hearing loss. The statistical data of National Institute on Deafness and Other Communication Disorders shows that 54% of the people over age 65 have hearing loss. Only 20% of the hearing loss wear hearing aids in America, which indicates many problems of hearing aids needed to be solved at present, such as speech enhancement, echo cancellation and automatic gain control. The aim of this study is to design and develop a PC-based hearing aid platform for hearing evaluation and prescription on the hearing loss.

　　This PC-based platform is developed to include pure tone audiogram (PTA) module and prescriptive gain fitting module by using MATLAB 6.5 and Simulink /DSP blockset software packages. Pure tone sound is generated by temporal sinusoidal sequences multiplying with various amplitudes in PTA module .With speech analysis-synthesis theory, the speech signal is sampled and framed by Hamming window and equalized by FIR filter-banks (low-pass, band-pass and high-pass filters) in prescriptive gain fitting module .

　　The evaluation and prescription platform for hearing aid are completely developed. There are pure tones of 250, 500, 1k, 2k, 4k, 8k and 16k Hz in PTA module to measure subject’s hearing threshold. The prescriptive gain fitting module provides 10 sub-bands speech equalizers which frequency range from 20~20k Hz, and the volume gain fitted according to gain fitting strategy with maximum volume at 105 dB SPL. The frequency calibration of the platform shows the relative error is under 0.1% .The volume mapping table of the platform is completed. The calibration results of volume output in the system relative to sound level meter shows the mean value of correlation coefficients is 0.976.

　　The phonetically balanced word list (PBWL) consists of 50 words is developed by analyzing the phonetically balanced Chinese sentences. According to the PBWL, the speech intelligibility(SI) test on Eight adults with normal hearing shows that the mean and standard deviation of SI score is 92 ± 2.33%.

　　The future work is suggested as follow: (1) frequency division: the band splitting based on multirate-signal processing is likely to decrease computational complexity and increase efficiency；(2) noise reduction: increasing SNR by using adaptive filter algorithms；(3) running simulation in Simulink/DSP blockset on PC can generate C codes from the model and down-load the C codes into the platform for real-time simulation. Finally, simulation algorithm can be implemented by DSP board.

[1]http://www.doh.gov.tw/, 行政院衛生署, 2003.
[2]http://volnet.moi.gov.tw/sowf/index.htm, 內政部社會司網站.
[3]Shu-Chen Pen , J.Brouce Tomblin, Hintat Cheung,&Lin-Sheue Wang,” Perception and Production of Mandrain Tones in Prelingully Deaf Children with cochlear Implants”.Ear & Hearing,2004.
[4]Jack Katz et al,“Handbook of clinical audiology” Williams & Wilkins ,1994.
[5]Xuedong Huang, Alex Acero and Hsiao-Wuen Hon ,”Spoken Language Processing” Prentice Hall PTR.
[6]余秀敏, 劉繼謚, ”國語語音特性平衡句之建立”, 電信研究季刊, 第19卷第1期, 民國78年3月.
[7]蕭雅文, “聽力學導論”, 五南出版社, 民國86年2月.
[8]N. Magotra, T. Hamill, B. Swartz, “Digital Signal Processing of Speech for the Hearing Impaired.” IEEE, In Proceedings of ASILOMAR-29, , 1996.
[9]林寶貴, ”聽覺障礙教育”, 五南出版社, 民國 83年.
[10]H.Gustav Mueller , James Hall, “Audiologists’ Desk Reference” Singular, 1998.
[11]王小川, ”語音訊號處理”,全華科技出版社.
[12]William F. Rintelmann, “Hearing Assessment”, pro-ed, 1991.
[13]Alan V. Oppenhein et al, ”Disctete-Time Signal Processing” Prentice Hall.1999.
[14]P.P. Vaidyanathan ,”Multirate systems and Filter Banks” Prentice Hall PTR.1993.
[15]鄭圳州,建構在PC上的耳鳴量測暨復健平台,國立成功大學醫學工程研究所碩士論文,2003.
[16]Saeed v. Vaseghi , “Advwnced Digital Signal Processing and Noise Reduction” John Wiley &Sons, 2000.
[17]蒙以正, ”數位信號處理應用MATLAB” ,旗標出版社, 2004.
[18]蒙以正, ”MATLAB入門與精進”, 儒林出社,2004.
[19]張智星, “MATLAB 程式設計”, 清蔚科技.
[20]羅華強, “訊號處理－MATLAB的運用”,全華科技出版社.
[21]Bruuce A .Dautrich, Lawrence ,Robiner,”On tthe Effects of Varying Filter Bank Parameters on Isolated Word Recognition,” IEEE Trans .Acous.,Speech, singal Processing, vol.ASSP-31, 4,August 1983,pp.793-806.
[22]Nathalie Virag,”Single Channel Speech Enhancement Based on Masking Properties of the Human Auditory,” IEEE Trans Speech, Audio Processing, vol.7, no2, March, 1999,pp.126-137.
[23]N. Magotra, T. Hamill, B. Swartz, “Digital Signal Processing of Speech for the Hearing Impaired.” IEEE, In Proceedings of ASILOMAR-29, 1996.
[24]H. Fletcher, ”Auditory Patterns Rev.Mod.Phy.,1940,12,pp47-65.
[25]Thomas F .Quatieri ,”Discrete-Time Speech signal Processing Principle and Practice”. Prentice Hall PTR.
[26]G. J.Tortora ,”Principle of human Anatomy” ,Happer Collins College Publishers,1997.
[27]A .Boothroyd , B.Mulhearn , J .Gong, and J .Ostroff , “Effect of spectral smearing of phoneme and word recognition,” J. Acoust. Soc. Am. 100(3), 1807-1818.
[28]http://www.nidcd.nih.gov/
[29]J.Punch, C. Chi , & J. Patterson,” A recommended protocol for prescriptive use of target gain rules”. Hearing Instruments, 41, 12, 14, 16, (1990).
[30]視覺化建模環境 Simulink 入門與進階，鈦思科技.
[31]王老得,張蓓莉,”中國語音均衡字彙表臨床應用之研討”,耳鼻喉科醫學會雜誌,第16卷第1號,1981.
[32]王老得,張蓓莉,”中國語音均衡字彙表之編制研究”,耳鼻喉科醫學會雜誌,第14卷第2號,1979.
[33]Sanjit K. Mitra, ”Digital Signal Processing”,McGeaw-Hill,2001