本研究目的是在可自動情境分類之雙麥克風除噪系統的適應性方向性麥克風策略中加入自動匹配系統，讓此除噪系統不會因為雙麥克風之間不匹配的問題，使除噪效能大大的降低。本研究先使用MATLAB(The MathWorks, Natick, Massachusetts, USA)軟體中的Simulink模擬軟體模擬雙麥克風在各種不匹配的狀況下，對於心型指向性產生的影響；並藉由本研究所發展的自動匹配演算法，在各種不匹配的狀況下能自動補償使得心型指向性成功的修正為理想的情況。接著將自動匹配演算法實現在TMS320C6713開發板(Texas Instruments, Dallas, Texas, USA)上，並與未匹配前的除噪系統進行比較。主觀評量方面經由HINT Pro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，實驗結果顯示加入自動匹配後的除噪系統經由自動情境分類控制後，使得SRT改善能夠有明顯提升的效果，而且比未加入自動匹配前的除噪系統有比較好的語音理解度結果。
語音品質的評估方面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(signal-to-noise ratio, SNR)超過15dB以上時，加入自動匹配的麥克風系統比未加入時獲得更好的語音品質，自動情境分類系統控制除噪策略開啟的PESQ指標有更低的失真影響。而在15dB以下時，加入自動匹配後的麥克風系統能夠準確地辨別噪音並降低自動情境分類系統造成的語音失真。由以上實驗結果驗證加入自動匹配系統能夠有效的提升語音理解度，並使得自動情境分類的結果更為準確。; Previously studies indicated that the adaptive directional microphone strategy has the characteristics of low computing cost and effective noise reduction. By tracking the noise source, this strategy could adaptively change the directivity of the directional dual-microphone to reduce the noise. However, when the two microphones were mismatched, the received signals showed differences on their phases and amplitudes. These differences would decrease the noise reduction performance if this mismatch was not well compensated.
The purpose of this study was to add an auto-matching process that matched the dual-microphones to improve the performance of automatic scene classification noise reduction system before the application of the adaptive directional microphone strategy. In this study, we first used Simulink (The MathWorks, Natick, Massachusetts, USA) to simulate the differences in polar directivity of cardioid were resulted from the varied conditions of microphone mismatch, and the auto-matching algorithms compensated the mismatch to achieve the ideal polar directivity of cardioid. Then, the auto-matching algorithms were implemented in TMS320C6713 DSP Starter Kit (Texas Instruments, Dallas, Texas, USA) and compared with the mismatched dual-microphone in the original noise reduction system. The speech reception thresholds (SRTs) from eight normal hearing subjects in different noise conditions were measured with the HINT Pro system (Bio-logic, Chicago, IL, USA) for subjective evaluation. The experimental results showed that the automatic scene classification noise reduction system provided significantly SRT effect and had better speech intelligibility.
The perceptual evaluation of speech quality (PESQ) was further used to estimate the quality of speech. Our experimental results showed that the auto-matching dual-microphone system provide more speech quality than those of the original dual-microphone system when the signal-to-noise ratio (SNR) is above 15dB. The PESQ index indicated less distortion of original signals with the auto-matching system. When the SNR is below 15dB, the auto-matching dual-microphone system could discriminate accurately the noise type and decrease the speech distortion caused by the automatic scene classification noise reduction system. The above-mentioned experimental results suggested that auto-matching system not only improve speech intelligibility but also let automatic scene classification noise reduction system obtain more accurate results.