運用TMS320C6713開發可自動情境分類之雙麥克風除噪系統

Development of an automatic scene classification noise reduction system with dual-microphone utilizing TMS320C6713

劉庭安(Ting-An Liu)

指導教授 ： 吳炤民

國立中央大學/資訊電機學院/電機工程學系/碩士(2012年)

摘要

本研究目的是以德州儀器TMS320C6713開發板(Texas Instruments, Dallas, Texas, USA)針對華語實現具有自動情境分類功能的適應性方向性麥克風除噪系統，此系統可在噪音環境下自動開啟麥克風除噪策略提升語音理解度，並在語音環境下關閉麥克風除噪策略保持語音品質。本研究使用的麥克風除噪策略為適應性方向性麥克風系統，此系統能根據噪音位置的變動適應性地改變系統的指向性達到除噪效果，實驗結果顯示此除噪系統能根據噪音位置的變動適應性地改變系統的指向性，最高具有約2.3dB的語音理解度加權式方向性指數(intelligibility-weighted directivity index, DISII)。而本研究使用多層感知機網路做為自動情境分類策略中的分類器並以八個與F0有關的特徵作為多層感知機網路的輸入，電腦模擬結果顯示此分類系統能達到平均97%的分類正確率，而實驗結果顯示此系統實作在TMS320C6713開發板上後能達到平均89.6%的分類正確率。本研究另外使用HINT Pro語言聽力檢查儀(Bio-logic, Chicago, IL, USA)對八位受測者在不同的噪音環境下進行語音接收閾值(speech reception threshold, SRT)的測試，結果顯示此系統能在噪音環境下降低最多6.2dB的平均SRT，而在語音品質的評估方面本研究使用語音品質客觀評量(perceptual evaluation of speech quality, PESQ)作為指標，實驗結果顯示，在訊噪比超過15dB時使用自動情境分類系統控制除噪策略開啟所得到的PESQ評分比不使用自動情境分類系統來的高，最大差距為0.24。由以上實驗結果可驗證此系統在噪音環境下能有效提升語音理解度，並在語音環境下保持語音品質不失真。

關鍵字

TMS320C6713 ； 噪音抑制 ； 情境分類 ； 適應性方向性麥克風

並列摘要

The purpose of this research was to develop an automatic scene classification noise reduction system with dual-microphone utilizing TMS320C6713 DSP Starter Kit (Texas Instruments, Dallas, Texas, USA). This system can automatically select the function of microphone noise reduction strategy to improve the intelligibility of speech in noise environment and turn off this function to maintain the quality of speech in speech environment. In this study, an adaptive directional microphone system was selected as microphone noise reduction strategy. Based on the noise direction, this system can adaptively change the directivity of the microphone to reduce the noise signal. The results showed that this system provides the function of adaptive change on system’s directivity and that the intelligibility-weighted directivity index (DISII) can reach 2.3dB. The multilayer perceptron network was used as the automatic scene classification strategy to classify speech or noise environment according to the eight F0-based features. The results of computer simulation and hardware implementation indicated that this system provides 97% average correct rate and reaches 89.6% average correct rate with TMS320C6713 DSP Starter Kit, respectively. Additionally, this study used HINT Pro system (Bio-logic, Chicago, IL, USA) to measure the speech reception threshold (SRT) from eight normal hearing subjects in different noise conditions. The results showed that this system can reduce 6.2dB SRT. The perceptual evaluation of speech quality (PESQ) was further used to estimate the quality of speech. Our experimental results showed that the difference of PESQ can reach up to 0.24 with and without using automatic scene classification strategy to control the noise reduction strategy. The above experimental results suggest that this system not only improve intelligibility of speech in noise environment but also keep the quality of speech in speech environment.

並列關鍵字

TMS320C6713 ； noise reduction ； automatic scene classification strategy ； adaptive directional microphone

參考文獻

Alexandre, E., Cuadra, L., Rosa, M. and Lopez-Ferreras, F. (2007). "Feature Selection for Sound Classification in Hearing Aids Through Restricted Search Driven by Genetic Algorithms." IEEE Transactions on Audio, Speech and Language Processing 15(8): 2249-2256.

ANSI (2007). Methods for Calculation of the Speech Intelligibility Index. ANSI S3.5-1997.

Bentler, R., Palmer, C. and Mueller, H. G. (2006). "Evaluation of a Second-Order Directional Microphone Hearing Aid: I. Speech Perception Outcomes." Journal of the American Academy of Audiology 17(3): 179-189.

Berghe, J. V. and Wouters, J. (1998). "An adaptive noise canceller for hearing aids using two nearby microphones." The Journal of the Acoustical Society of America 103(6): 3621-3626.

Buchler, M., Allegro, S., Launer, S. and Dillier, N. (2005). "Sound classification in hearing aids inspired by auditory scene analysis." EURASIP Journal on Applied Signal Processing 2005: 2991-3002.