【摘要】 根据阵列信号语音增强的思想,提出一种基于频域处理的谱相减与波束形成相结合的语音增强结构。结构为多路信号输入,每路含噪信号在谱相减后,增加了波束形成结构,不仅有效地消除了背景噪声,也抑制了谱相减后的音乐噪声。并使用该算法对实际环境中采集到的含噪语音信号进行了仿真,结果显示经过该系统处理后的增强语音的信噪比有了较大的提高,主观试听效果也很好。更多还原

【Abstract】 A speech enhancement algorithm combining spectral subtraction and beamforming is proposed in this paper,which can reduce the residual noise effectively with a microphone array.The speech signal is extracted from the noisy signal after the final adaptive filtering due to the novel structure of the microphone array.Simulation to the noisy speech of real environment show that the system works efficiently even when the signal to noise ratio(SNR)is very low.更多还原