摘要
语音识别系统易受噪声影响,采用谱减法等传统语音增强算法滤除噪声,但存在“音乐噪声”的困扰。针对该类问题,提出一种基于分数阶傅里叶与三次样条插值的语音增强算法。对语音进行分数阶傅里叶变换,并采用谱减法对含噪语音进行预处理,通过基于维纳滤波的噪声估计器来估计噪声,实现噪声的迭代更新;采用三次样条插值将含噪语音与估计噪声进行函数化;将函数化后的含噪语音与估计噪声进行几何谱减法处理,获得纯净语音。仿真结果显示:与传统的语音增强算法相比,所提算法在改善“音乐噪声”的问题上效果明显,同时也改善了语音可懂性和语音质量。
The speech recognition system is susceptible to noise.In order to filter noise,traditional speech enhancement methods such as spectral subtraction are often used by researchers,but these methods are troubled by“music noise”.To solve this problem,a speech enhancement algorithm based on cubic spline interpolation is proposed in this paper.Firstly,speech is subjected to the fractional Fourier transform,spectral subtraction is employed to preprocess noisy speech,and the Wiener filter noise estimator is utilized to realize the iterative update of noise.Secondly,the speech with noise and the estimated noise is functioned by cubic spline interpolation.A geometric spectrum subtraction algorithm is then used to process the functional speech with noise and the estimated noise to produce the pure speech.The simulation results show that compared with the traditional speech noise reduction algorithm,the proposed algorithm has an obvious effect on improving the“music noise”problem,and also greatly improves speech intelligibility and speech quality.
作者
周坤
陈文杰
陈伟海
林岩
孙先涛
ZHOU Kun;CHEN Wenjie;CHEN Weihai;LIN Yan;SUN Xiantao(School of Electrical Engineering and Automation,Anhui University,Hefei 230601,China;Anhui Engineering Laboratory of Human-Robot Collaboration System and Intelligent Equipment,Anhui University,Hefei 230601,China;School of Automation Science and Electrical Engineering,Beihang University,Beijing 100191,China)
出处
《北京航空航天大学学报》
EI
CAS
CSCD
北大核心
2023年第10期2826-2834,共9页
Journal of Beijing University of Aeronautics and Astronautics
基金
国家自然科学基金(51975002)。
关键词
语音增强
三次样条插值
维纳滤波
几何谱减法
智能控制
speech enhancement
cubic spline interpolation
Wiener filter
geometric spectrum subtraction algorithm
intelligent control
