CAJ | 학술논문

　　针对传统小波阈值去噪法中硬阈值函数不连续，软阈值函数有固定偏差的缺点，提出一种新的阈值函数。对Donoho的固定阈值进行改进，提出一种自适应的阈值。在Matlab环境中，分别进行了实验选取最优小波基，新阈值函数的最优参数以及新阈值函数与传统硬阈值函数，软阈值函数和折衷阈值函数的对比。实验结果表明，新的阈值函数能更有效地提高语音信号的信噪比，改进语音质量。
　　침대전통소파역치거조법중경역치함수불련속，연역치함수유고정편차적결점，제출일충신적역치함수。대Donoho적고정역치진행개진，제출일충자괄응적역치。재Matlab배경중，분별진행료실험선취최우소파기，신역치함수적최우삼수이급신역치함수여전통경역치함수，연역치함수화절충역치함수적대비。실험결과표명，신적역치함수능경유효지제고어음신호적신조비，개진어음질량。
To overcome the disadvantage in traditional wavelet threshold de-noising approach, the hard threshold function is discontinuous while the soft threshold function has fixed deviation;a new threshold function is proposed. An adaptive threshold is proposed to enhance the fixed threshold put forward by Donoho. Under the Matlab environment, three experiments are separately carried to select the optimum wavelet base, the optimum parameters of the new threshold function and to compare the new threshold function with traditional threshold functions such as the hard threshold function, the soft threshold function and the compromise threshold function. Experimental result shows that the new threshold function is more effective to enhance the Signal-to-Noise Ratio and to improve the quality of the de-noised speech signal.