Comparison and Improvement for Fine Resolution Frequency Estimation from DFT Samples

• Main Authors: Hsin-Lan Chi, 紀馨嵐
• Other Authors: Jan-Ray Liao
• Format: Others
• Language: zh-TW
• Published: 2012
• Online Access: http://ndltd.ncl.edu.tw/handle/54521263728355712956

碩士 === 國立中興大學 === 通訊工程研究所 === 100 === 　　The parameter estimation of complex sinusoidal waveform under white noise usually consists of two stages. The first stage is the coarse search which searches the maximum magnitude in the N-point discrete Fourier transform (DFT) from an input of N samples. The second stage is the fine search which searches around the peak determined in the first stage. 　　To date, the Candan method is the best method for fine search using DFT coefficient interpolation. It uses three DFT coefficients to achieve high resolution frequency estimation. It is derived from Jacobsen’s estimation equation and added a bias correction term to increase its accuracy. The correction term is effective for the high SNR and adds almost no additional computational cost. Therefore, it can be used in all SNR levels. 　　This thesis presents three methods that reduce the number of DFT coefficients used in the estimation and three methods that use the same number of samples as in Candan method. In the experiments, the biases and the root mean square (RMS) errors of the six methods are compared with those of Candan method. We show that our methods can be better than Candan method and both the bias and RMS error can be improved significantly.