In dynamic spectrum access, a new wireless system (secondary user: SU) can utilize the spectrum licensed to an existing wireless system (primary user: PU) while the spectrum is vacant. For accomplishing effective and reliable spectrum utilization, statistics of the spectrum usage by the PU, such as duty cycle (DC), are useful for the SU. In this paper, we investigate accurate spectrum measurement technique with energy detector for obtaining the accurate duty cycle (DC) estimation with a noise floor (NF) estimation which is used for proper threshold setting. In one conventional method, median filtered FCME (Forward Consecutive Mean Excision), frequency selectivity due to interference among spectra of independent symbols degrades the NF and DC estimation performances. In this paper, we propose Welch FFT based NF estimation and DC estimation since Welch FFT has a potential to suppress the effect of the frequency selectivity. We analytically obtain probability density function (PDF) and right tail probability of output of the Welch FFT process and this analysis enables to set proper threshold. Numerical results will show that the proposed method can achieve better NF estimation performance. In addition, the NF estimation improvement and the suppression of frequency selectivity by the Welch FFT can provide around 15 dB SNR gain in the DC estimation performance.