Timely identification of fetuses with the risk of asphyxia during labor enables clinicians to prevent potential adverse outcomes. The fetal welfare assessment is usually based on the analysis of fetal heart rate (FHR) and uterine contraction (UC) signals obtained through the Cardiotocograph (CTG). The CTG analysis involves visual interpretation of different morphological signal patterns making use of proposed medical guidelines. Unfortunately, this methodology has been shown to lack objectivity leading to a poor interpretation reproducibility. According to the literature, each fetus has its own control behavior and the risk to evolve from one health condition to another frequently changes during the labor process. Likewise, experienced clinicians attempt to consider the temporal evolution when interpreting CTGs. In this context, several approaches based on time-varying signal processing techniques have been proposed. However, they are mainly focused on simple fetal reactivity as a response to a UC, without taking into account the progression of the frequency dynamical changes occurring over time. In this work, we propose to analyze such progression along the FHR signal by using empirical mode decomposition (EMD) and time-varying spectral-based analysis. The main idea is to analyze if a particular FHR episode in the time-domain reflects characteristic spectral changes in the frequency-domain that can be related to the fetal distress. For this purpose, first the FHR signal is decomposed by using EMD. Then, for each EMD component the time-varying spectrum is computed in order to study its frequency behavior over time. The proposed method has been tested on real CTG data extracted from the open access CTU-UHB database. Results show that each FHR deceleration episode leaves a particular spectral signature in the frequency components associated with the neural sympathetic fetal reactivity (0.03-0.15Hz), whose progression over time could help to assess the fetal condition during labor.