探索双通道定位算法与盲源分离相结合进行噪声分离的可能。根据双通道空间定位理论，以声音传播过程中在双通道间产生的延时差和强度差为线索对混合信号进行分离。当待分离的辐射噪声中存在时频混叠或同一位置多个源时，双通道定位分离算法能分离出不同位置处源信号，然后利用盲源分离方法对同一位置处的多个源信号进行再分离。在Roomsim中进行分离方案的可行性验证，假设房间六个壁能100%吸收所有频率信号，不存在反射与混响，对比分析声源位置及个数对分离结果的影响。结果表明，当源位置的个数不大于3时，双通道定位分离算法能有效分离出不同位置的声源信号，用盲源分离方法可进一步分离出相同位置处的不同源信号，但当源位置个数大于等于4时双通道定位分离算法的分离性能降低。%The noise source identification based on dual-channel localization algorithm and blind source separation (BSS) is studied. According to the principle of the dual-channel localization algorithm, the noise sources in different places can be separated based on time-delay difference and intensity difference between the two channels in sound propagation process. When the radiation noise has a time-frequency overlap or multiple excitation sources appear in a same place, the dual-channel localization algorithm can separate the source signals in different places. Then, the BSS method is adopted to re-separate the multiple source signals at the same place. Finally, the feasibility of this method is verified in Roomsim, and the influence of source number and location on separation performance is analyzed. The results show that the separation method combining the dual-channel localization algorithm with BSS can separate the sources efficiently and is independent of the noise from the other positions. However, the dual-channel localization algorithm does not work well when the source number is larger than 3.
展开▼
