声级残差最小二乘非线性拟合法估值混响时间

Evaluation of Reverberation Time Based on a Sound Level Residual Error Least-Squares Nonlinear Regression Approach

提出了估值混响时间的声级残差最小二乘非线性拟合法.该方法充分利用了脉冲响应从早期到后期的所有测量数据,降低了信噪比和积分上限对混响时间估值结果的影响.同时,对比现有其他方法在统计上有最小的估值标准差,其均值的95%概率置信区间也最小,有效地提高了混响时间的估值精度.克服了脉冲响应积分法测量混响时间中,由于较低的脉冲响应信噪比或积分上限选择的不当所产生的估值偏差.

The paper proposes a nonlinear regression approach with sound level residual error leastsquares for evaluating reverberation times, when reverberation decay curves are obtained by the backward integration of room impulse responses. The analysis shows that this method makes full use of the measured data of the impulse response from start to finish. The influences of the signal-to-noise ratio and the integration limit on reverberation times evaluation are weakened, and a minimum statistical standard deviation and 95 % probability confidence interval are obtained. The deviated evaluation caused by either the low signal-to-noise ratio or the careless choice of the integration limit is avoided, and a precisive reverberation time evaluation is achieved.