基于环形存储结构的非分段式FFT算法  

Non-segmented FFT algorithm based on circular storage structure

作　　者：薛雷[1] 李长远[1] 杜敏[1] 

机构地区：[1]上海大学通信与信息工程学院,上海200072

出　　处：《电子测量技术》2010年第12期35-37,63,共4页Electronic Measurement Technology

摘　　要：常用的分段式FFT算法响应时间长,硬件资源要求高,而且在模式识别系统设计中分段式FFT算法容易丢失特征值,影响识别精度。针对连续语音识别(鼾声识别)系统设计中特征值的提取提出了一种基于环形存储结构的非分段式流水FFT算法,并在理论上证明了该算法的正确性,通过对实际语音识别系统的测试表明非分段式流水FFT算法不仅耗费的硬件资源比分段式FFT算法低,识别精度高,并且能够连续实时地工作,保证整个系统的实时性!Segmented FFT algorithm commonly takes long response time and demands hardware resource,as well as in the pattern recognition system design segmented FFT algorithm is easy to lose characteristic values,which affect the recognition accuracy.In this paper,we present a non-segmented pipelined FFT algorithm based on ring storage structure for the design of continuous speech recognition system eigenvalue extraction.we prove the correctness of the algorithm in theory,and by means of practical tests in the peech recognition system show that the non-segmented flow FFT algorithm not only consume lower hardware resources than segmented FFT algorithm,possess high identification accuracy,but also work continuously and real-time to ensure the real-time of the whole system.

关 键 词：环形缓冲区 FFT算法 

分 类 号：TP31[自动化与计算机技术—计算机软件与理论]