Recently, there has been increasing interest in applying graph theory to the quantitative analysis of brain functional networks, while phase synchronization (PS) analysis has been demonstrated to be a useful method to infer functional connectivity with multichannel neural signals, e.g., electroencephalogram (EEG). In this paper, we focus on the case that the number of channels in EEG data is not adequate for the use of graph theory analysis. The degree of network-links (DNLs), an index based on the PS analysis of all the EEG wave pairs, is proposed to study the relevant and the overall characteristics of the brain. With the help of a novel division to the frequency range 0.5–30 Hz, we analyze the DNLs in different frequency bands of the EEG signals. As a comparison, a frequency band analysis of the relative power spectrum is conducted. The results demonstrate that when the cerebral infarction (CI) patients and normal control people are analyzed, there is a need for the reasonable length of EEG data to quantify the differences between different dynamical systems; under a reasonable data length, the frequency band (19–24 Hz) yields the best accuracy for diagnosing CI, which lies within the classical beta band (13–30 Hz); furthermore, only in the 19–24 Hz band, as for the values of relative power spectrum, in each EEG channel, there presents a similar relationship between the CI group and control group. The experimental results suggest that 19–24 Hz should be the optimal range for the diagnosis of CI, further the DNLs calculated within this band serve as an assist indicator in the CI diagnosis.