Objective: In the context of fetal heart rate (fHR) estimation, the article addresses the fusion of two monomodal estimations into a multimodal one. Electrical and mechanical modalities are considered through the use of abdominal electrocardiogram (ECG) and phonocardiogram (PCG). The aim of the fusion is to provide a fHR estimation, robust against noise sources, and especially against the risk of confusion with the mother heart rate (mHR).Approach: A hidden Markov chain is considered to model variations of the two monomodal fHR estimations and the true fHR. Thanks to the Viterbi's algorithm (VA), the two fHR estimations are then merged. However, the classical VA does not ensure that the merged estimation always follows the true fHR: it may alternate between the fHR and the mHR, especially when the mother component is not correctly removed from the abdominal ECG. Therefore, a modified VA is proposed to efficiently avoid confusion with mHR. The aim is to discourage large variations between successive states.Results: Comparisons between classical and modified VA are performed on real pregnant women data. The modified VA reduces the confusion between fHR and mHR in major cases compared to the classical VA. For most recordings, the confusion with mHR decreases under 1%, and for the best case, it is reduced from 59% to 0%.Conclusion: The modified VA succeeds to improve the fHR estimation by reducing confusion with mHR. Significance: Fusion of estimations from two modalities is a promising approach for more robust fHR monitoring.