Abstract: Since the number of signals needs to be determined when we perform direction of arrival (DOA) estimation algorithms, wideband signals have to be sampled many times in every frequency for traditional methods, otherwise, it can't be judged correctly.To this question, the paper proposes a new method for determining the number of wideband signals based on Bootstrap criterion.First, the wideband signals are divided into several non-overlapping subbands, eigen decomposition is performed to every part of the data, and the eigenvectors are weighted by corresponding eigenvalues in every frequency.Furthermore, Bootstrap criterion is employed to construct resampling matrix, then the number of signals in these frequencies is estimated by some clustering techniques.Finally, the results of every frequency are integrated to derive the number of wideband signals.The method avoids the process of focusing, and it has a high probability of success under the circumstance of non-Gaussian noise, especially when the snapshots is small.The performance is proved by simulation results at last.