The complexity of systems with parallel acquisition architectures has increased sharply with the continuous growth of signal bandwidth, and sampling precision has been greatly challenged. Considering that the signal to be measured usually exhibits sparsity in the frequency domain, this paper proposes a re-configurable bandwidth interleaved acquisition architecture to maximize test flexibility and accuracy. The sampling process is divided into two stages: sensing and re-configurable acquisition. In the sensing stage, the signal spectrum distribution is roughly understood using a sparse Fourier transform. Based on the spectrum sensing results, the subband selection and the adaptive adjustment of the mixing local oscillator in the acquisition system are performed. Ultimately, these steps enhance sampling accuracy. This paper verifies the effectiveness of this method on a 10 GHz acquisition system, demonstrating that it can significantly reduce data redundancy. In addition, it improves acquisition accuracy compared to traditional bandwidth interleaved systems. The experimental results indicate that re-configurable sampling can significantly improve the quality of sampling results. This is evidenced by a signal-to-noise ratio improvement of over 7.4 dB and a spurious-free dynamic range improvement of over 4.7 dB compared to traditional sampling results.