<p>The successful launch of the Cyclone Global Navigation Satellite System (CYGNSS) has opened an unprecedented opportunity for rapid observation of Wind Speed (WS) across vast oceanic regions. However, considerable debate persists over the choice of input feature parameters for WS retrieval models based on CYGNSS data, and enhancing the accuracy of WS retrieval is a focal point of current research. To address the aforementioned problems, this study establishes a comprehensive CYGNSS wind speed retrieval feature parameter set through an in-depth analysis of CYGNSS data, thereby providing a reference and basis for selecting input features for WS retrieval models. Through this analysis, we identified three crucial observational features: the normalized bistatic radar cross section, leading edge slope, and signal-to-noise ratio. Using these features, we developed a WS retrieval model based on the geophysical model function for CYGNSS data. Furthermore, acknowledging the intrinsic interconnection between wind and wave dynamics, we incorporate significant wave height into the WS retrieval model to further improve the WS retrieval accuracy. Comparative assessments with datasets from the European Centre for Medium-Range Weather Forecasts, the Chinese-French Oceanography Satellite Scatterometer, and buoy WS data underscore the high accuracy of our model, demonstrating its utility as a valuable tool for research in ocean dynamics and marine environmental prediction.</p>