摘要
针对加性有色噪声,提出了语音信号模糊消噪算法;建立并训练了一个语音模糊消噪系统——自适应神经模糊推理系统(ANFIS);用其对含噪语音中的有色噪声进行模糊估计,从而提取出干净的语音。对算法进行了仿真实验,结果表明,对模拟有色噪声在-17dB时能提取出清晰的语音。
Aiming at additive colored noise, this paper presents a new fuzzy denoising algorithm of speech signal based on the principle of fuzzy inference in fuzzy mathematics. A fuzzy denoising system of speech signal Auto Neural Fuzzy Inference System （ANFIS） is set up and trained with MATLAB. Colored noise can be successfully removed by using subtraction in the original speech signal, and a fuzzy estimation of colored noise in the speech signal which con- tains noises can be made for extracting clean speech signal. An emulation experiment on the algorithm for different signal to noise ratio has been done, and the results show that the system can extract clean speech signal from the simulated colored noise at -17 dB of signal to noise ratio. This proves that the algorithm is very effective.
出处
《声学技术》
CSCD
2009年第5期682-685,共4页
Technical Acoustics
关键词
语音信号
有色噪声
白噪声
模糊消噪
自适应神经模糊推理系统
speech signal
colored noise
white noise
fuzzy denoise
Auto Neural Fuzzy Inference System （ANFIS）
