Research and development on signal separation techniques which can work effectively for interfaces

本研究项目的目的是提出一种从移动源和海的混合物中回收移动源和海的分离方法。对于分离运动源的方法，我们提出了三种方法：1.分离运动源的方法。一种特征向量算法，一种超指数方法，一种3a波束形成算法。特征向量算法可以在一次迭代中得到恢复源所需的解。超指数方法可以使恢复源的解以指数速率得到期望的解。在波束形成方面，首先对运动源进行定位，然后根据定位信息对运动源进行波束形成恢复。在三种方法中，第三种方法对分离运动源效果最好。对于分离刘海的方法，我们认为使用第三种方法可以恢复刘海。关于什么是适合人机接口的方法，我们将继续研究它的解决方案。

The objective of the research project is to propose the separation method for recovering moving sources and bangs from their mixtures. As for the method for separating moving sources, we have proposed three kinds of methods : 1. An eigenvector algorithms, 2 A super-exponential method, and 3 A beamforming. The eigenvector algorithm can work such that the desired solution for recovering sources can be obtained with one iteration. The super-exponential method can work so that the desired solution for recovering sources can be obtained with exponential rate. On the beamforming, at first, the localization of moving sources is implemented and then based on the information on the localization, the beamforming is applied to recover the moving sources. Among the three proposed method, the third one is best for separating moving sources. As for the method for separating bangs, we consider that the bang can be recovered by using the third method. On the problem of what are suitable methods for the interface between human and machine, we will continue to investigate its solutions.