论文部分内容阅读
人工神经网络实现技术的研究具有重大意义，它可推动神经网络研究的进一步深化并加快其实用化步伐。本文重点研究了基于ＤＳＰ（数字信号处理）的神经网络虚拟实现技术，设计并实现了以ＴＭＳ３２０Ｃ３０作为处理器的并行处理系统，峰值速度为１亿ＦＬＯＰＳ，并行效率可达到９０％（ＢＰ回忆阶段）。该系统可以方便地扩展为更大规模的系统，用于水声信号处理，以满足实际应用的要求。
Artificial neural network to achieve technology research is of great significance, it can promote the further deepening of neural network research and accelerate its practical pace. This paper focuses on the virtual realization technology of neural network based on DSP (digital signal processing), designs and realizes the parallel processing system with TMS320C30 as processor, the peak speed is 100 million FLOPS, the parallel efficiency can reach 90% (BP recall stage) . The system can be easily extended to larger systems for underwater acoustic signal processing to meet the requirements of practical applications.