With more increased interest in renewable energy for reducing air pollution and accomplishing low or zero emission, the polymer electrolyte membrane fuel cell (PEMFC) has been recognized as the most attractive and promising candidate with the merits of low-operating temperature, fast start-up ability, and high efficiency, etc. Specifically, for an efficient operation of the PEMFC, it is absolutely required to obtain the fuel cell management system (FCMS). Among several factors in the FCMS, the state-of-health (SOH) diagnostic methodology is more critically considered for reliability and durability of the PEMFC. For this purpose, an experimental output voltage signal (EOVS) that adequately depicts a single cell and fuel cell stack behavior is used as fundamental information. Then, the most significant task for achieving an improved SOH diagnosis is to gain an elaborate EOVS. Unfortunately, because of the use of unexpected and instantaneous sensing of noise-riding EOVS in the FCMS, it is unavoidable that non-corrected EOVS surely results in erroneous SOH. Consequently, it is strictly required to implement an innovative research that can find a solution to aforementioned problem. Thus, this research newly gives insight to the design and implementation of the discrete wavelet transform (DWT)-based noise reduction in the noise-riding EOVS, namely denoising technique. At the DWT in this research, the noise-riding EOVS can be assumed as an original signal with non-stationary and transient phenomena according to the multi-resolution analysis (MRA) with a function of both time and frequency localization. The detailed procedure for denoising the noise-riding EOVS is as follows. First, through the MRA, it is possible to decompose the noise-riding EOVS into two sub-bands of low- and high-frequency components, A_n and D_n. Second, to adjust the detailed coefficient in high-frequency component, two denoising methods of hard- and soft-thresholding with VisuShrink-based threshold value are implemented so as to clearly separate the signal from the noise. Finally, the inverse DWT is used to reconstruct the de-noised EOVS using an adjusted detailed coefficient. Experimental results definitely showed the robustness of this proposed approach. Consequently, it can be sufficiently expected to have an optimal solution for noise reduction of the EOVS and to lead an improved FCMS. For reference, an experimental apparatus was basically designed for obtaining the noise-riding EOVS from the PEMFC by the ‘Materials and Electro-Chemistry Laboratory in Inha University’.