摘要
从计算机模拟和１４Ｎ远程ＮＱＲ实验上研究了神经网络在对已知信号的自动识别问题．证实了多层网络的反向传播算法对解决这一问题的有效性．实验结果表明．在信噪比很低和存在很强的干扰条件下，Ｂ－Ｐ算法仍然保持了很高的识别率（≥９０％）．这在远程ＮＱＲ实际应用中是非常有意义的．
Problems associated to signal automatic recognition by Neural Networks is studied with computer simulations and 14N remote NQR experiments.The effectiveness of the multiple layer back propagation algorithm for solving the problems is venfied.The experimental result indicates that,under the condition of very low signal-to-noise ratio and very strong interferences,the B-P algorithm still keeps a very high recognition (>90%).This is of significant improtance in remote NQR applications.
出处
《波谱学杂志》
CAS
CSCD
1996年第3期283-290,共8页
Chinese Journal of Magnetic Resonance
基金
上海市科委"青年科技启明星重点培养计划"资助
关键词
信号识别
神经网络
核四极共振
Signal Recognition, Neural networks, Remote NQR
