The improvement of symbol rate estimation by the wavelet transform

The symbol rate estimator by the wavelet transform proposed in Ho et al., and Chan et al. (1997) suffered from difficulties of choosing wavelet scales and the weak ability of anti-noise. In this paper we proposed an improved version of such an estimator. The main modification of our algorithm is that we move the carrier frequency of the received waveform to zero before the wavelet transform, i.e. we extract the symbol rate from the WT magnitudes of the baseband modulated signal instead of those of the MF signals. We find that with this modification there will be no invalid wavelet scales as occurred in Ho and choosing large scale not only reduces the additive noise, but also enhances the strength of the desired spike series, which remarkably improve the anti-noise ability of the estimator. Simulation results verified the effectiveness of our algorithm.