在源信号和传输信道未知情况下，只利用接收天线的观测数据抽取源信号，称为盲信号分离.盲信号分离不仅是信号处理界、而且也是神经网络界的研究热点课题，在无线数据通信、雷达、图像、语音、医学以及地震信号处理等领域都具有广阔的应用前景.采用自然梯度法和分阶段学习法。-The source signal and transmission channel is unknown circumstances, using only observational data extraction receiving antenna source signal, known as blind signal separation. Blind signal separation is not only the signal processing community, but also a neural network hot topic in research circles, in the wireless data communications , radar images, audio, medical and seismic signal processing and other fields have broad application prospects. and the phases using the natural gradient learning method.