摘要
提出了一种基于小波变换和卡尔曼滤波相结合的语音增强方法,这样既保留了小波变换对自相似过程的去相关作用和多分辨分析的功能,同时又保持了卡尔曼滤波器对未知信号的线性无偏最小方差估计的特点,可以有效地减小非平稳噪声;并引入基于声学模型的感知滤波器,以提高语音信号的可懂度。实验证明该方法对于低信噪比的有色噪声干扰条件下的语音信号的增强效果要优于一般的语音增强系统。
An approach based on wavelet and Kalman filter for speech enhancement is proposed,thus to retain the characters of multi-resolution analysis in wavelet domain and decorrelation in self-similarity,process while to maintain the linear unbiased and minimum error variance estimation of Kalman filtering for the unknown signal.This has effectively reduced the non-stationary noise,and by concatenating the perceptual filter based on psychoacoustic model with the Kalman filter,could further improve the speech intelligibility.Simulation results show that the proposed approach has the best performance in reducing noise with low SNR in colored noise environments.
出处
《通信技术》
2010年第4期152-154,共3页
Communications Technology
关键词
语音增强
小波变换
卡尔曼滤波
听觉模型
speech enhancement
wavelet transform
Kalman filter
psychoacoustic model
