公开(公告)号：KR100844176B1

公开(公告)日：2008-07-07

申请号：KR1020080007904

申请日：2008-01-25

Applicant: 재단법인서울대학교산학협력재단

Inventor： 김남수 ,  이승열

IPC: G10L21/0208 ,  G10L15/20 ,  H04M9/08

Abstract: A method for RES based on a statistical model is provided to determine a current signal state with one of four assumption divided according to the presence of an NET(Near End Talker) signal and a residual echo, estimate the PSD of associated signals based on the assumption for the determined current signal state, acquire the optimal gain of an RES filter by using the estimated PSD of the signals, and configure the optimal RES filter, thereby effectively suppressing a residual echo. A method for RES(Residual Echo Suppression) based on a statistical model comprises the following steps of: determining a current signal state with one of four assumption divided according to the presence of an NET signal and a residual echo(S10); estimating a background noise PSD(Power Spectrum Density) and a residual echo PSD on the basis of the assumption for the determined current signal state(S20); estimating an SNR(Signal To Noise Ratio) and an SER(Signal To Echo Ratio) by using the estimated background noise PSD, the residual echo PSD, and the spectrum components of an AEC(Acoustic Echo Canceller) output signal(S30); estimating a pre-SNR and a pre-SER by using the estimated post-SNR and post-SER, the gain of a residual echo cancellation filter of a prior frame, and the spectrum components of the AEC output signal of the prior frame(S40); acquiring the gain of the residual echo cancellation filter by using the estimated pre-SNR and pre-SER(S50); and applying the acquired gain of the residual echo cancellation filter to the residual echo cancellation filter to cancel the residual echo(S60).

Abstract translation: 提供了一种基于统计模型的RES的方法，以根据NET（近端对讲机）信号和残留回波的存在而划分的四个假设之一来确定当前信号状态，基于相关信号估计相关信号的PSD 假设确定的当前信号状态，通过使用信号的估计PSD获取RES滤波器的最佳增益，并且配置最佳RES滤波器，从而有效地抑制残留回波。 基于统计模型的RES（残余回波抑制）的方法包括以下步骤：根据NET信号和残留回波的存在分为四个假设之一确定当前信号状态（S10）; 基于所确定的当前信号状态的假设来估计背景噪声PSD（功率谱密度）和残余回波PSD（S20）; 通过使用估计的背景噪声PSD，残差回波PSD和AEC（声学回波消除器）输出信号的频谱分量来估计SNR（信噪比）和SER（信号到回波比）; 通过使用估计的后SNR和后SER估计前SNR和前SER，前一帧的残余回声消除滤波器的增益和先前帧的AEC输出信号的频谱分量（S40 ）; 通过使用估计的前SNR和前SER获取残余回声消除滤波器的增益（S50）; 以及将所获取的残余回声消除滤波器的增益应用于残差回声消除滤波器以消除残留回波（S60）。