It is worthwhile noting that during this work we also examined the STSA estimator which minimizes (1) under the additional assumption that the signal is not surely present in the noisy observation [ 11, [3]. While this estimator results in a further reduction of the residual noise in comparison with that obtained by using (20), it also introduces an effect of low-pass filtering on the enhanced speech signal. This effect is reduced as the assumed probability of signal absence is lowered; but then the amount of residual noise reduction gained by this estimator is also reduced. For the above reasons we found it unworthy to incorporate the signal presence uncertainty in the log STSA estimator.翻译成中文

值得注意的是，在这项工作中，我们还研究了在附加假设下最小化(1)的STSA估计器，即在嘈杂的观测中信号不一定存在[11，[3]。虽然这个估计器相对于使用(20)得到的残余噪声进一步减少，但它也会在增强语音信号上引入低通滤波的效果。随着假设的信号不存在的概率降低，这种效应会降低；但是，这个估计器获得的残余噪声降低量也会降低。出于以上原因，我们认为将信号存在不确定性纳入对数STSA估计器是不值得的。