In the field of vibration monitoring, RFID tags offer advantages such as low cost, easy deployment, and compact size. More importantly, they can be effectively deployed in locations where conventional sensors are difficult or impractical to install-such as in confined spaces, or under line-of-sight obstructions. However, traditional vibration sensing methods based on the single tag are limited by low sampling rates and limited interference resistance, making them insufficient for high-precision sensing. This study proposes a novel vibration frequency sensing system via passive RFID tags that fuses the phases of multiple tags under nonuniform sampling to enable robust and accurate frequency sensing. Firstly, a novel theoretical model for vibration sensing via multiple tags is built, revealing the relationship between phase signals and vibration signals. Then, an alternating sampling mechanism among multiple tags is used to improve the effective sampling rate and the signal-to-noise ratio (SNR). To address spectral distortion caused by non-uniform sampling, interpolation-based resampling and spectral masking algorithms are designed, thereby reducing frequency sensing errors. Finally, extensive experimental results demonstrate that, compared to traditional sensing methods based on a single tag, the proposed phase fusion algorithm of multiple tags achieves an average SNR improvement of approximately 52%, a reduction in absolute frequency error by about 63%, and a decrease in relative error of nearly 67%. This enables high-precision, low-error vibration detection, and providing a new solution for low-cost, wireless vibration sensing.