This paper proposes a carrier recovery method tailored to narrow-band multicarrier backscatter wireless links, which employ subcarrier modulation. Because backscattered subcarrier is usually produced by low accuracy and low rate internal clock in a low power consumption or batteryless, wireless sensor, the corresponding receiver needs to compensate the subcarrier frequency drift and fluctuation dynamically for demodulation and decoding. Although a backscattered subcarrier is produced by changing the antenna reflection coefficient of wireless sensor, the backscattered signal is usually converted to a zero-crossing signal to enhance signal resolution, and edge detection based demodulator is applied at the receiver. However, in a multicarrier backscatter system, the edges at symbol boundaries of each backscattered subcarrier become indistinct because of the narrow bandwidth. Consequently, subcarrier frequency estimation with existing carrier recovery technique may results in error. This paper solves the problem by removing outliers of temporal subcarrier frequency based on the statistical distribution and applying a feedback control. The statistical carrier recovery method is examined experimentally with a prototype wireless sensor and a software defined receiver to reveal the method can automatically compensate unstable subcarrier with about 10-2 order drift and outperforms the popular frequency doubler based carrier recovery. The optimal feedback gain, 1.0, is also revealed.