It is important to obtain reliable attenuation results from experimental data to elucidate the physical mechanism responsible for ultrasonic wave attenuation. For attenuation estimation, a time window is often used to compute the frequencies of the direct-arrival waveforms. However, the effect of windowing distorts the spectral distribution due to a spectral leakage effect, degrading the attenuation estimates. We propose a method that enables accurate measurement of ultrasonic attenuation using sweep signals under the assumptions that velocity dispersion can be ignored and the quality factor [Formula: see text] is not dependent on frequency. We obtained the spectral amplitude of the sweep signal in the frequency-time domain using the continuous wavelet transform and estimated attenuation in the time-scale spectrum domain using the spectral-ratio method. This method is independent of the effect of windowing, whereas the windowing effect underestimates the attenuation results. In the absence of noise, the estimated attenuation results using sweep signals are in perfect agreement with the given input values, and the accuracy of the estimated attenuation results from windowed pulse waveforms depends on the extraction window length. However, our numerical experiments demonstrated that the proposed method is largely influenced by the existence of overlapping sweep events such as multiple reflections between the source and receiver transducer. Thus, applicability of the proposed method is limited to highly attenuative media, in which overlapping events are much smaller than direct sweep signals because these multiple reflected events are largely attenuated. Application of the proposed method to laboratory experimental data yielded similar underestimation of the attenuation results due to the windowing effect in the case of highly attenuative media. We also evaluated the usefulness of observing compressed pulse waveforms with shorter duration from the crosscorrelation of sweep waveforms than the case of pulse generation.