摘要
针对高速实时信号处理的需要,提出了一种对任意长度序列进行FFT的快速改进算法。通过对FFT处理前数据添零个数和DFT分解参数的优化选择,显著降低了FFT处理的运算量。结合频域脉冲压缩等信号处理实例,探讨了该算法在高速DSP上实现时的资源分配、程序编程以及传输I/O瓶颈问题,分别提出了具体的解决方法,并在实际DSP系统中测试了这种改进算法的性能指标,将其和普通算法的性能作了比较。
An improvement of FFT algorithm for long sequences is proposed, aiming to reduce the huge computation burden in high speed real time signal processing. By optimizing the number of padding zeros and the DFT decomposition parameters, the computation volume of FFT is decreased significantly. This algorithm is implemented on the high speed TS101 DSP for application to pulse compression processing. By analyzing resources allocation,programming and I/O bottleneck, methods are suggested to improve the algorithm′s performance. The performance is demonstrated in real time signal processing and is compared with ordinary processing method.
出处
《现代雷达》
CSCD
北大核心
2005年第7期23-26,共4页
Modern Radar
