To address the recognition challenge of Multi-h Continuous Phase Modulation (Multi-h CPM) signals with varying modulation parameters, this paper proposes a modulation recognition algorithm grounded in fuzzy entropy theory. This theory transcends the binary approach of distance and count-based similarity in approximate entropy, opting for a membership function to assess similarity and more accurately reflect the complexity of time series. The algorithm separates and calculates the fuzzy entropy of the in-phase and quadrature components of the received signal, utilizing these values as classification features for a Support Vector Machine(SVM). Experiments demonstrate that the algorithm achieves 100% recognition accuracy for full-response rectangular shaped Multi-h CPM signals across various modulation index sets at signal-to-noise ratios above 6 dB, and enables modulation recognition with a minimal number of symbols.