Abstract: Based on Time-Frequency（TF） ratio Blind Sources Separation（BSS） algorithm proposed by M.Puigt and Y.Deville，the paper proposes a new BSS approach via adopting multi-resolution of S transform.Firstly，the approach uses S transform to obtain TF plane of mixed signals and constructes different TF ratio matrices.Secondly，it detectes single source occurs to identify each element of the mixing matrix，and then obtaines the estimated signals.This approach can separate non-stationary sources effectively and has the multi-resolution characteristics.

Key words: S transform, time-frequency ratio, blind source separation, nonstationary signals

摘要： 在M.Puigt和Y.Deville提出的时频盲源分离算法基础上，引入S变换来获取非平稳信号的多分辨率特性。首先通过S变换将一维混叠信号映射到二维时频平面，然后构造不同混叠信号的时频比矩阵，通过在时频比矩阵范围内搜索单源分析域计算混合阵的每个元素，进而估计源信号。该方法能有效分离非平稳信号且具备多分辨率特性。

关键词: S变换, 时频比, 盲源分离, 非平稳信号