In view of the possibility of false peaks on the surface electromyography (sEMG) of patients with spasticity, leading to decreased signal differences before and after stretch reflex onset (SRO), a method for detecting SRO based on empirical mode decomposition (EMD) denoising and modified sample entropy recognition is proposed. First, the EMG signal is decomposed via EMD. Then, the soft threshold is set to denoise the decomposed signal on the basis of the sEMG signal of the subjects in resting state. Lastly, modified sample entropy is used to identify SRO. The experimental results show that the EMD algorithm can effectively remove noise from the EMG signal, and the average recognition rate of SRO under the optimal parameter of the modified sample entropy is 94%.