- 學位論文
頻譜亂度在行動電話語音品質上的應用
The Spectral Entropy Application on Voice Quality of Handset
若您是本文的作者，可授權文章由華藝線上圖書館中協助推廣。
摘要
行動電話對於通話中語音的品質，並沒有客觀的標準可以評量。目前在業界有兩種方式可以獲得客觀的結果，一則是透過更多的工程師測試聆聽的主觀比較，另一個方式是購買其他儀器設備。第一種方式容易流於主觀；而第二種方式則需要額外的成本資出。這篇論文裡，以目前一般公司已有的儀器設備，設計可以量化語音品質的方式。 在一段語音信號中，加入片段雜訊或是片段靜音，用來模擬移動通話時最常見的兩種情況：背景噪音太大，以及聲音斷斷續續。接著使用音頻分析儀，分析頻譜亂度。實驗結果顯示，加入雜訊或是靜音的語音訊號，頻譜亂度是增加的。 接著利用這項結果，搭配基地台模擬器，模擬實際的應用。測試幾家不同品牌的行動電話，客觀的比較結果，與一般實際使用的語音品質主觀感覺相同。 對於語音在吵雜環境的判斷，提出更接近實際情況的解決方式。文獻上，模擬背景噪音，多半採用兩種格式，一是白雜訊，另一是粉紅雜訊。但這兩種雜訊與實際的背景雜音有出入。在這本論文裡，測試幾種吵雜環境，利用頻譜亂度，定義一些背景噪音的規則，利用這些規則，來做語音雜訊的處理。
並列摘要
There is no standard to get voice quality in this field. There are two ways to get voice quality from mobile phone, one is to listen the voice quality by many engineers, and the other is to purchase related equipment to help in getting voice quality. However, both ways have disadvantages, the first way is too subjective and the second way is too expensive. The complains about the mobile voice quality are mainly divided into two, one is voice with background noise, the other is chopping voice. If we want to simulate these situations, we can add noise and silence to a speech source separately. After Audio Analyzer, both of the voices with noise and chopping have higher Entropy than original speech source. In some mobile voice quality benchmark tests, the subjective result is the same entropy analysis with equipment. It means that the voice quality can be ranked objectivity. The background noise studies are very important to reduce background noise during voice communication. Most of papers use white and pink as their background noises. After measuring real environment noise, the white and pink noises are not very popular at all, so we provide a rule to decide background noise according Spectral Entropy.
參考文獻
[4] Bing-Fei Wu and Kun-Ching Wang,”Noise Spectrum Estimation with Entropy-Based VAD in Non-stationary Environments”, IEICE Trans. Fundamentals, vol.E89–A, no.2, 2006
[5] Chuan JIA and Bo XU, “An Improved Entropy-based Endpoint Detection Algorithm”, Institute of Automation, 2002
[6] Aik Ming Toh and Roberto Togneri, “Spectral Entropy as Speech Features for Speech Recognition”, ICASSP 83, Boston, 2005
[13] Doh-Suk Kim, “An Auditory Model for Single-Ended Speech Quality Estimation”, IEEE Trans. on Speech and Audio Processing, vol. 13, no. 5, 2005
[14] David M.Howard and Jamie Angus, “Acoustic and Psychoacoustics”, 3rd edition, Focal Press, ISBN:0-240-51995-7, pp.26