Signal decomposition into the frequency components is one of the oldest challenges in the digital signal processing. In early nineteenth century, Fourier transform (FT) showed that any applicable signal can be decomposed by unlimited sinusoids. However, the relationship between time and frequency is lost under using FT. According to many researches for appropriate time-frequency representation, in early twentieth century, wavelet transform (WT) was proposed. WT is a well-known method which developed in order to decompose a signal into frequency components. In contrast with original WT which is not adaptive according to the input signal, empirical wavelet transform (EWT) was proposed. In this paper, the performance of discrete WT (DWT) and EWT in terms of signal decomposing into basic components are compared. For this purpose, a stationary signal including five sinusoids and ECG as biomedical and nonstationary signal are used. Due to being non-adaptive, DWT may remove signal components but EWT because of being adaptive is appropriate. EWT can also extract the baseline of ECG signal easier than DWT.