请提供上下文或学科范围,否则没法猜.呵呵，原句是：instead of using the FFT-derived estimates of the noisy speech and noise spectra directly in the attenuation rule,we use smoothed versions of the power spectra. 文章是关于含噪语音信号的噪声抑制方法的，请指教！谢谢FFT（快速傅立叶变换）的衍生换算法。