Abstract: This paper presents an improved speech enhancement method, this method decomposes the noisy speech signal into sub-band signals, then filters the sub-band signals using the adaptive filter in Discrete Fractional Cosine Transform（DFRCT）which has good orthogonal properties, and adaptive filter uses Least Mean Square（LMS） algorithm. It obtains the enhanced sub-band speech signals by using inverse discrete fractional cosine transform to the filtered signal. It composes all enhanced sub-bands signal. Simulation experiments indicate that the proposed algorithm reduces the computing time about 10 s and possesses good performance both in SegSNR and PESQ tests.

Key words: Discrete Fractional Cosine Transform（DFRCT）, filter banks, adaptive filter, sub-band decomposition, speech enhancement

摘要： 提出一种改进的语音增强方法，将带噪语音信号进行子带分解，再对子带信号进行离散分数余弦变换（DFRCT）域滤波，利用了DFRCT良好的正交特性，且自适应滤波采用最小均方（LMS）算法。对滤波后的信号进行DFRCT逆变换得到增强后的子带语音信号，合成增强后的语音信号。仿真结果表明，该算法在减少输入信号自相关程度的基础上，提高了收敛速度，减少了计算时间（约10 s），增强后的语音信号的分段信噪比（SegSNR）和PESQ值都有所提高，具有良好的语音增强效果。

关键词: 离散分数余弦变换, 滤波器组, 自适应滤波, 子带分解, 语音增强