Aiming at the difficulty of power grid supraharmonic sampling and data processing, this paper proposes a supraharmonic signal detection method based on the windowed stochastic demodulation architecture. Firstly, the method utilizes analog information conversion technology combined with the idea of adding windows to obtain the compressed sampled data, which is realized by compressing and sampling the supraharmonic signal at the sampling end. Subsequently, the step-by-step bisection method is adopted to reconstruct the sampled data quickly, which is realized by setting the first-order differentiation of the residuals as the convergence threshold to improve the sparsity adaptive matching tracking (SAMP) algorithm. Simulation and test results show that the method can detect the frequency amplitude of supraharmonic signals accurately with only 10% of the data. The frequency amplitude detection error is less than 2% , and the frequency detection error is less than 3 Hz. The method provides a new way for the supraharmonic measurement.