摘要
针对短波通信语音信号中噪声在频域分布的特点,引入了基于数据驱动的噪声估计算法,改进了频点信噪比的计算,提出了基于Bark子带的多种增强算法联合处理的方法。实验结果表明,该算法能有效抑制短波语音信号中的噪声,处理后的语音失真度小,语音自然度较高,避免了残留音乐噪声。
This study boosts the performance of speech enhancement algorithm in high-frequency communication by leveraging the characteristic distribution in frequency domain. A data-driven noise estimation algorithm is discussed and Bark frequency band SNR is calculated, with multiple enhancement technigues presented. Simulation results indicate obvious speech distortion reduction, with little musical noise retained.
出处
《信息工程大学学报》
2012年第1期60-65,65,共6页
Journal of Information Engineering University
关键词
短波通信
语音增强
噪声估计
Bark子带
high-frequency communication
speech enhancement
noise estimation
Bark band
