一种基于最大似然的混响时间盲估计方法*

混响是室内声学中的重要现象,在室内设计与音频信号处理中都需要测量或估计混响时间。本文改进了一种基于最大似然估计的混响时间盲估计方法,即采用说话人在房间中自然说话时发出的混响语音信号来估计混响时间的方法。该方法首先确定语音衰减段的最优边界,其次计算该衰减段的两个额外参数,据此筛选出符合条件的语音段,最后将满足条件的语音段采用最大似然估计得到混响时间估计值。在五个不同混响时间条件下的仿真表明,与已有方法相比,改进方法估计的混响时间同真实混响时间的偏差更小,方差更低,估计准确性较高。

An improved algorithm for blind estimation of reverberation time based on maximum likelihood*

Reverberation is a well-known phenomenon in the field of room acoustics and reverberation time is an important parameter to be measured or estimated in acoustic design and audio signal processing. An improved algorithm for maximum likelihood based blind estimation of reverberation time is presented in this paper, i.e., using the reverberant speech signal naturally uttered by the speaker to estimate reverberation time. Firstly special rules for defining the optimal boundaries of the speech decay segments are introduced, then two additional parameters are calculated to select the speech decay segments for maximum likelihood estimation, finally, the reverberation time is estimated based on the maximum-likelihood estimation using the selected speech decay segments. Simulations under five different reverberant conditions show that both the deviation and their variance between the estimated reverberation time and the true value are decreased.